Training in progress, step 50, checkpoint

last-checkpoint/adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:72698aa029c702a244945ef40f373c767a14c49bb90cb2368b40a8bd023d1bbd
+oid sha256:d6754747b9ef7b418555c7a60f6d7224496e5b588b5c9c729b5bbb355306e1b2
 size 34456

last-checkpoint/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b6c477eaec40cd6ac98dbd68a0a4b41efb85c6e3e415c1d310eaacdda2cc138e
+oid sha256:720367065ae9dc88bd08c6476d06387891d584bc15ef3262b9979b19eeaa80f6
 size 73222

last-checkpoint/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f96fd81130ea63001ae86aaba8896cde979c1f9e3d91c8d71a7955a135db6962
+oid sha256:881608cbb07448770a5e76289b30723e66073325ba435fcdc41b961bd5ae0ba0
 size 14244

last-checkpoint/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c47edf82128c3f034f21204d4b9ce5c76cd3269748ce31061b148a0a389d049d
+oid sha256:1479f554a7a167fd87e9ce99235e9d065370465268e2f8373a74df4bd50982ce
 size 1064

last-checkpoint/trainer_state.json

@@ -1,16 +1,16 @@
 {
   "best_metric": 11.5,
   "best_model_checkpoint": "miner_id_24/checkpoint-50",
-  "epoch": 0.007275637527738368,
+  "epoch": 0.001818909381934592,
   "eval_steps": 50,
-  "global_step": 200,
+  "global_step": 50,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 3.637818763869184e-05,
-      "grad_norm": 3.0213206628104672e-05,
+      "grad_norm": 2.948543624370359e-05,
       "learning_rate": 1.0060000000000002e-05,
       "loss": 23.0,
       "step": 1
@@ -18,350 +18,350 @@
     {
       "epoch": 3.637818763869184e-05,
       "eval_loss": 11.5,
-      "eval_runtime": 151.3209,
-      "eval_samples_per_second": 76.493,
-      "eval_steps_per_second": 19.125,
+      "eval_runtime": 151.3578,
+      "eval_samples_per_second": 76.474,
+      "eval_steps_per_second": 19.12,
       "step": 1
     },
     {
       "epoch": 7.275637527738368e-05,
-      "grad_norm": 2.2829284716863185e-05,
+      "grad_norm": 2.2105969037511386e-05,
       "learning_rate": 2.0120000000000004e-05,
       "loss": 23.0,
       "step": 2
     },
     {
       "epoch": 0.00010913456291607552,
-      "grad_norm": 1.8357066437602043e-05,
+      "grad_norm": 1.7058524463209324e-05,
       "learning_rate": 3.018e-05,
       "loss": 23.0,
       "step": 3
     },
     {
       "epoch": 0.00014551275055476736,
-      "grad_norm": 2.194116859755013e-05,
+      "grad_norm": 2.060344922938384e-05,
       "learning_rate": 4.024000000000001e-05,
       "loss": 23.0,
       "step": 4
     },
     {
       "epoch": 0.0001818909381934592,
-      "grad_norm": 2.0882815078948624e-05,
+      "grad_norm": 2.0100153051316738e-05,
       "learning_rate": 5.03e-05,
       "loss": 23.0,
       "step": 5
     },
     {
       "epoch": 0.00021826912583215103,
-      "grad_norm": 3.217526318621822e-05,
+      "grad_norm": 3.034624933206942e-05,
       "learning_rate": 6.036e-05,
       "loss": 23.0,
       "step": 6
     },
     {
       "epoch": 0.00025464731347084286,
-      "grad_norm": 2.8971715437364765e-05,
+      "grad_norm": 2.815291918523144e-05,
       "learning_rate": 7.042e-05,
       "loss": 23.0,
       "step": 7
     },
     {
       "epoch": 0.0002910255011095347,
-      "grad_norm": 1.676463034527842e-05,
+      "grad_norm": 1.748102658893913e-05,
       "learning_rate": 8.048000000000002e-05,
       "loss": 23.0,
       "step": 8
     },
     {
       "epoch": 0.00032740368874822657,
-      "grad_norm": 3.206745532224886e-05,
+      "grad_norm": 3.207884219591506e-05,
       "learning_rate": 9.054000000000001e-05,
       "loss": 23.0,
       "step": 9
     },
     {
       "epoch": 0.0003637818763869184,
-      "grad_norm": 2.8715619919239543e-05,
+      "grad_norm": 2.7205978767597117e-05,
       "learning_rate": 0.0001006,
       "loss": 23.0,
       "step": 10
     },
     {
       "epoch": 0.00040016006402561027,
-      "grad_norm": 1.8758235455607064e-05,
+      "grad_norm": 1.864074874902144e-05,
       "learning_rate": 0.00010007052631578948,
       "loss": 23.0,
       "step": 11
     },
     {
       "epoch": 0.00043653825166430207,
-      "grad_norm": 2.735985253821127e-05,
+      "grad_norm": 2.7616073566605337e-05,
       "learning_rate": 9.954105263157895e-05,
       "loss": 23.0,
       "step": 12
     },
     {
       "epoch": 0.0004729164393029939,
-      "grad_norm": 4.027846807730384e-05,
+      "grad_norm": 3.965870928368531e-05,
       "learning_rate": 9.901157894736842e-05,
       "loss": 23.0,
       "step": 13
     },
     {
       "epoch": 0.0005092946269416857,
-      "grad_norm": 4.429101318237372e-05,
+      "grad_norm": 4.447407991392538e-05,
       "learning_rate": 9.84821052631579e-05,
       "loss": 23.0,
       "step": 14
     },
     {
       "epoch": 0.0005456728145803776,
-      "grad_norm": 3.3590822567930445e-05,
+      "grad_norm": 3.3188109227921814e-05,
       "learning_rate": 9.795263157894737e-05,
       "loss": 23.0,
       "step": 15
     },
     {
       "epoch": 0.0005820510022190694,
-      "grad_norm": 2.6291587346349843e-05,
+      "grad_norm": 2.72152046818519e-05,
       "learning_rate": 9.742315789473686e-05,
       "loss": 23.0,
       "step": 16
     },
     {
       "epoch": 0.0006184291898577613,
-      "grad_norm": 4.353782787802629e-05,
+      "grad_norm": 4.269916098564863e-05,
       "learning_rate": 9.689368421052633e-05,
       "loss": 23.0,
       "step": 17
     },
     {
       "epoch": 0.0006548073774964531,
-      "grad_norm": 7.02256802469492e-05,
+      "grad_norm": 7.144509436329827e-05,
       "learning_rate": 9.63642105263158e-05,
       "loss": 23.0,
       "step": 18
     },
     {
       "epoch": 0.0006911855651351449,
-      "grad_norm": 6.216230394784361e-05,
+      "grad_norm": 5.8125104260398075e-05,
       "learning_rate": 9.583473684210527e-05,
       "loss": 23.0,
       "step": 19
     },
     {
       "epoch": 0.0007275637527738368,
-      "grad_norm": 5.73032557440456e-05,
+      "grad_norm": 5.5581836932105944e-05,
       "learning_rate": 9.530526315789474e-05,
       "loss": 23.0,
       "step": 20
     },
     {
       "epoch": 0.0007639419404125286,
-      "grad_norm": 5.283307473291643e-05,
+      "grad_norm": 4.992326284991577e-05,
       "learning_rate": 9.477578947368422e-05,
       "loss": 23.0,
       "step": 21
     },
     {
       "epoch": 0.0008003201280512205,
-      "grad_norm": 3.86905885534361e-05,
+      "grad_norm": 3.754639328690246e-05,
       "learning_rate": 9.424631578947369e-05,
       "loss": 23.0,
       "step": 22
     },
     {
       "epoch": 0.0008366983156899123,
-      "grad_norm": 4.2620362364687026e-05,
+      "grad_norm": 4.4129676098236814e-05,
       "learning_rate": 9.371684210526316e-05,
       "loss": 23.0,
       "step": 23
     },
     {
       "epoch": 0.0008730765033286041,
-      "grad_norm": 2.9412009098450653e-05,
+      "grad_norm": 3.2032086892286316e-05,
       "learning_rate": 9.318736842105263e-05,
       "loss": 23.0,
       "step": 24
     },
     {
       "epoch": 0.000909454690967296,
-      "grad_norm": 7.737339910818264e-05,
+      "grad_norm": 7.238514808705077e-05,
       "learning_rate": 9.26578947368421e-05,
       "loss": 23.0,
       "step": 25
     },
     {
       "epoch": 0.0009458328786059878,
-      "grad_norm": 6.505755300167948e-05,
+      "grad_norm": 5.7584187743486837e-05,
       "learning_rate": 9.212842105263159e-05,
       "loss": 23.0,
       "step": 26
     },
     {
       "epoch": 0.0009822110662446796,
-      "grad_norm": 7.259017729666084e-05,
+      "grad_norm": 7.019707845756784e-05,
       "learning_rate": 9.159894736842107e-05,
       "loss": 23.0,
       "step": 27
     },
     {
       "epoch": 0.0010185892538833714,
-      "grad_norm": 5.836385753354989e-05,
+      "grad_norm": 5.2452742238529027e-05,
       "learning_rate": 9.106947368421054e-05,
       "loss": 23.0,
       "step": 28
     },
     {
       "epoch": 0.0010549674415220635,
-      "grad_norm": 0.00012386597518343478,
+      "grad_norm": 0.00011238550359848887,
       "learning_rate": 9.054000000000001e-05,
       "loss": 23.0,
       "step": 29
     },
     {
       "epoch": 0.0010913456291607553,
-      "grad_norm": 0.0001556278148200363,
+      "grad_norm": 0.00013983648386783898,
       "learning_rate": 9.001052631578948e-05,
       "loss": 23.0,
       "step": 30
     },
     {
       "epoch": 0.001127723816799447,
-      "grad_norm": 7.392914994852617e-05,
+      "grad_norm": 6.872925587231293e-05,
       "learning_rate": 8.948105263157895e-05,
       "loss": 23.0,
       "step": 31
     },
     {
       "epoch": 0.0011641020044381389,
-      "grad_norm": 6.028249117662199e-05,
+      "grad_norm": 5.7285513321403414e-05,
       "learning_rate": 8.895157894736842e-05,
       "loss": 23.0,
       "step": 32
     },
     {
       "epoch": 0.0012004801920768306,
-      "grad_norm": 0.0001456019381294027,
+      "grad_norm": 0.00013172421313356608,
       "learning_rate": 8.842210526315789e-05,
       "loss": 23.0,
       "step": 33
     },
     {
       "epoch": 0.0012368583797155227,
-      "grad_norm": 8.436777716269717e-05,
+      "grad_norm": 8.472959598293528e-05,
       "learning_rate": 8.789263157894738e-05,
       "loss": 23.0,
       "step": 34
     },
     {
       "epoch": 0.0012732365673542145,
-      "grad_norm": 9.687324927654117e-05,
+      "grad_norm": 9.260691149393097e-05,
       "learning_rate": 8.736315789473685e-05,
       "loss": 23.0,
       "step": 35
     },
     {
       "epoch": 0.0013096147549929063,
-      "grad_norm": 0.00013006821973249316,
+      "grad_norm": 0.00011570382775971666,
       "learning_rate": 8.683368421052632e-05,
       "loss": 23.0,
       "step": 36
     },
     {
       "epoch": 0.001345992942631598,
-      "grad_norm": 0.00012605867232196033,
+      "grad_norm": 0.00013264735753182322,
       "learning_rate": 8.63042105263158e-05,
       "loss": 23.0,
       "step": 37
     },
     {
       "epoch": 0.0013823711302702899,
-      "grad_norm": 8.003330731298774e-05,
+      "grad_norm": 7.428869866998866e-05,
       "learning_rate": 8.577473684210527e-05,
       "loss": 23.0,
       "step": 38
     },
     {
       "epoch": 0.0014187493179089819,
-      "grad_norm": 9.521541505819187e-05,
+      "grad_norm": 9.110313112614676e-05,
       "learning_rate": 8.524526315789474e-05,
       "loss": 23.0,
       "step": 39
     },
     {
       "epoch": 0.0014551275055476737,
-      "grad_norm": 0.00017305008077528328,
+      "grad_norm": 0.00014309283869806677,
       "learning_rate": 8.471578947368421e-05,
       "loss": 23.0,
       "step": 40
     },
     {
       "epoch": 0.0014915056931863655,
-      "grad_norm": 0.00016906786186154932,
+      "grad_norm": 0.00016003995551727712,
       "learning_rate": 8.41863157894737e-05,
       "loss": 23.0,
       "step": 41
     },
     {
       "epoch": 0.0015278838808250573,
-      "grad_norm": 0.0001861914643086493,
+      "grad_norm": 0.00016799979493953288,
       "learning_rate": 8.365684210526317e-05,
       "loss": 23.0,
       "step": 42
     },
     {
       "epoch": 0.001564262068463749,
-      "grad_norm": 0.00020132421923335642,
+      "grad_norm": 0.00016025554214138538,
       "learning_rate": 8.312736842105264e-05,
       "loss": 23.0,
       "step": 43
     },
     {
       "epoch": 0.001600640256102441,
-      "grad_norm": 0.0002141977456631139,
+      "grad_norm": 0.000203637799131684,
       "learning_rate": 8.259789473684211e-05,
       "loss": 23.0,
       "step": 44
     },
     {
       "epoch": 0.0016370184437411329,
-      "grad_norm": 0.0001226340827997774,
+      "grad_norm": 9.858429984888062e-05,
       "learning_rate": 8.206842105263158e-05,
       "loss": 23.0,
       "step": 45
     },
     {
       "epoch": 0.0016733966313798247,
-      "grad_norm": 0.0001583473349455744,
+      "grad_norm": 0.00014126356109045446,
       "learning_rate": 8.153894736842105e-05,
       "loss": 23.0,
       "step": 46
     },
     {
       "epoch": 0.0017097748190185165,
-      "grad_norm": 0.0002799605717882514,
+      "grad_norm": 0.00022886847727932036,
       "learning_rate": 8.100947368421053e-05,
       "loss": 23.0,
       "step": 47
     },
     {
       "epoch": 0.0017461530066572083,
-      "grad_norm": 0.0002967856707982719,
+      "grad_norm": 0.00025842548348009586,
       "learning_rate": 8.048000000000002e-05,
       "loss": 23.0,
       "step": 48
     },
     {
       "epoch": 0.0017825311942959,
-      "grad_norm": 0.00010255781671730801,
+      "grad_norm": 0.00010498856863705441,
       "learning_rate": 7.995052631578949e-05,
       "loss": 23.0,
       "step": 49
     },
     {
       "epoch": 0.001818909381934592,
-      "grad_norm": 0.0003075671847909689,
+      "grad_norm": 0.0002834200859069824,
       "learning_rate": 7.942105263157896e-05,
       "loss": 23.0,
       "step": 50
@@ -369,1084 +369,10 @@
     {
       "epoch": 0.001818909381934592,
       "eval_loss": 11.5,
-      "eval_runtime": 151.2138,
-      "eval_samples_per_second": 76.547,
-      "eval_steps_per_second": 19.138,
+      "eval_runtime": 151.2778,
+      "eval_samples_per_second": 76.515,
+      "eval_steps_per_second": 19.13,
       "step": 50
-    },
-    {
-      "epoch": 0.0018552875695732839,
-      "grad_norm": 6.767702871002257e-05,
-      "learning_rate": 7.889157894736843e-05,
-      "loss": 23.0,
-      "step": 51
-    },
-    {
-      "epoch": 0.0018916657572119757,
-      "grad_norm": 6.940264574950561e-05,
-      "learning_rate": 7.83621052631579e-05,
-      "loss": 23.0,
-      "step": 52
-    },
-    {
-      "epoch": 0.0019280439448506675,
-      "grad_norm": 0.00013540219515562057,
-      "learning_rate": 7.783263157894737e-05,
-      "loss": 23.0,
-      "step": 53
-    },
-    {
-      "epoch": 0.0019644221324893593,
-      "grad_norm": 0.00010234087676508352,
-      "learning_rate": 7.730315789473684e-05,
-      "loss": 23.0,
-      "step": 54
-    },
-    {
-      "epoch": 0.0020008003201280513,
-      "grad_norm": 3.912624379154295e-05,
-      "learning_rate": 7.677368421052632e-05,
-      "loss": 23.0,
-      "step": 55
-    },
-    {
-      "epoch": 0.002037178507766743,
-      "grad_norm": 0.0001325135672232136,
-      "learning_rate": 7.624421052631579e-05,
-      "loss": 23.0,
-      "step": 56
-    },
-    {
-      "epoch": 0.002073556695405435,
-      "grad_norm": 0.0001106501804315485,
-      "learning_rate": 7.571473684210526e-05,
-      "loss": 23.0,
-      "step": 57
-    },
-    {
-      "epoch": 0.002109934883044127,
-      "grad_norm": 0.00014796988398302346,
-      "learning_rate": 7.518526315789475e-05,
-      "loss": 23.0,
-      "step": 58
-    },
-    {
-      "epoch": 0.0021463130706828185,
-      "grad_norm": 0.00020090723410248756,
-      "learning_rate": 7.465578947368422e-05,
-      "loss": 23.0,
-      "step": 59
-    },
-    {
-      "epoch": 0.0021826912583215105,
-      "grad_norm": 6.893830868648365e-05,
-      "learning_rate": 7.412631578947369e-05,
-      "loss": 23.0,
-      "step": 60
-    },
-    {
-      "epoch": 0.002219069445960202,
-      "grad_norm": 0.00012792616325896233,
-      "learning_rate": 7.359684210526317e-05,
-      "loss": 23.0,
-      "step": 61
-    },
-    {
-      "epoch": 0.002255447633598894,
-      "grad_norm": 0.00010460198973305523,
-      "learning_rate": 7.306736842105264e-05,
-      "loss": 23.0,
-      "step": 62
-    },
-    {
-      "epoch": 0.002291825821237586,
-      "grad_norm": 8.348002302227542e-05,
-      "learning_rate": 7.253789473684211e-05,
-      "loss": 23.0,
-      "step": 63
-    },
-    {
-      "epoch": 0.0023282040088762777,
-      "grad_norm": 0.00018085635383613408,
-      "learning_rate": 7.200842105263158e-05,
-      "loss": 23.0,
-      "step": 64
-    },
-    {
-      "epoch": 0.0023645821965149697,
-      "grad_norm": 0.000223978262511082,
-      "learning_rate": 7.147894736842105e-05,
-      "loss": 23.0,
-      "step": 65
-    },
-    {
-      "epoch": 0.0024009603841536613,
-      "grad_norm": 0.00012863126175943762,
-      "learning_rate": 7.094947368421052e-05,
-      "loss": 23.0,
-      "step": 66
-    },
-    {
-      "epoch": 0.0024373385717923533,
-      "grad_norm": 0.0001924559473991394,
-      "learning_rate": 7.042e-05,
-      "loss": 23.0,
-      "step": 67
-    },
-    {
-      "epoch": 0.0024737167594310453,
-      "grad_norm": 9.336372750112787e-05,
-      "learning_rate": 6.989052631578948e-05,
-      "loss": 23.0,
-      "step": 68
-    },
-    {
-      "epoch": 0.002510094947069737,
-      "grad_norm": 0.00024296657647937536,
-      "learning_rate": 6.936105263157896e-05,
-      "loss": 23.0,
-      "step": 69
-    },
-    {
-      "epoch": 0.002546473134708429,
-      "grad_norm": 0.0002156701812054962,
-      "learning_rate": 6.883157894736843e-05,
-      "loss": 23.0,
-      "step": 70
-    },
-    {
-      "epoch": 0.0025828513223471205,
-      "grad_norm": 0.0001783580519258976,
-      "learning_rate": 6.83021052631579e-05,
-      "loss": 23.0,
-      "step": 71
-    },
-    {
-      "epoch": 0.0026192295099858125,
-      "grad_norm": 0.000139903902891092,
-      "learning_rate": 6.777263157894737e-05,
-      "loss": 23.0,
-      "step": 72
-    },
-    {
-      "epoch": 0.0026556076976245045,
-      "grad_norm": 0.00022315837850328535,
-      "learning_rate": 6.724315789473684e-05,
-      "loss": 23.0,
-      "step": 73
-    },
-    {
-      "epoch": 0.002691985885263196,
-      "grad_norm": 0.00027320714434608817,
-      "learning_rate": 6.671368421052631e-05,
-      "loss": 23.0,
-      "step": 74
-    },
-    {
-      "epoch": 0.002728364072901888,
-      "grad_norm": 0.0003263742255512625,
-      "learning_rate": 6.61842105263158e-05,
-      "loss": 23.0,
-      "step": 75
-    },
-    {
-      "epoch": 0.0027647422605405797,
-      "grad_norm": 0.00010844551434274763,
-      "learning_rate": 6.565473684210527e-05,
-      "loss": 23.0,
-      "step": 76
-    },
-    {
-      "epoch": 0.0028011204481792717,
-      "grad_norm": 0.00021539705630857497,
-      "learning_rate": 6.512526315789474e-05,
-      "loss": 23.0,
-      "step": 77
-    },
-    {
-      "epoch": 0.0028374986358179637,
-      "grad_norm": 0.00018285549595020711,
-      "learning_rate": 6.459578947368421e-05,
-      "loss": 23.0,
-      "step": 78
-    },
-    {
-      "epoch": 0.0028738768234566553,
-      "grad_norm": 0.0002819342480506748,
-      "learning_rate": 6.406631578947369e-05,
-      "loss": 23.0,
-      "step": 79
-    },
-    {
-      "epoch": 0.0029102550110953473,
-      "grad_norm": 0.0002171331288991496,
-      "learning_rate": 6.353684210526316e-05,
-      "loss": 23.0,
-      "step": 80
-    },
-    {
-      "epoch": 0.002946633198734039,
-      "grad_norm": 0.0004205996810924262,
-      "learning_rate": 6.300736842105263e-05,
-      "loss": 23.0,
-      "step": 81
-    },
-    {
-      "epoch": 0.002983011386372731,
-      "grad_norm": 0.00041594888898544014,
-      "learning_rate": 6.247789473684212e-05,
-      "loss": 23.0,
-      "step": 82
-    },
-    {
-      "epoch": 0.003019389574011423,
-      "grad_norm": 0.00038625358138233423,
-      "learning_rate": 6.194842105263159e-05,
-      "loss": 23.0,
-      "step": 83
-    },
-    {
-      "epoch": 0.0030557677616501145,
-      "grad_norm": 0.00024285879044327885,
-      "learning_rate": 6.141894736842106e-05,
-      "loss": 23.0,
-      "step": 84
-    },
-    {
-      "epoch": 0.0030921459492888066,
-      "grad_norm": 0.00017769836995285004,
-      "learning_rate": 6.088947368421053e-05,
-      "loss": 23.0,
-      "step": 85
-    },
-    {
-      "epoch": 0.003128524136927498,
-      "grad_norm": 0.00035690178629010916,
-      "learning_rate": 6.036e-05,
-      "loss": 23.0,
-      "step": 86
-    },
-    {
-      "epoch": 0.00316490232456619,
-      "grad_norm": 0.00021868004114367068,
-      "learning_rate": 5.9830526315789475e-05,
-      "loss": 23.0,
-      "step": 87
-    },
-    {
-      "epoch": 0.003201280512204882,
-      "grad_norm": 0.00040930655086413026,
-      "learning_rate": 5.9301052631578946e-05,
-      "loss": 23.0,
-      "step": 88
-    },
-    {
-      "epoch": 0.0032376586998435737,
-      "grad_norm": 0.0005343824159353971,
-      "learning_rate": 5.877157894736843e-05,
-      "loss": 23.0,
-      "step": 89
-    },
-    {
-      "epoch": 0.0032740368874822658,
-      "grad_norm": 0.00020195181423332542,
-      "learning_rate": 5.82421052631579e-05,
-      "loss": 23.0,
-      "step": 90
-    },
-    {
-      "epoch": 0.0033104150751209573,
-      "grad_norm": 0.00029612492653541267,
-      "learning_rate": 5.771263157894737e-05,
-      "loss": 23.0,
-      "step": 91
-    },
-    {
-      "epoch": 0.0033467932627596494,
-      "grad_norm": 0.00038756342837587,
-      "learning_rate": 5.718315789473685e-05,
-      "loss": 23.0,
-      "step": 92
-    },
-    {
-      "epoch": 0.003383171450398341,
-      "grad_norm": 0.00035572919296100736,
-      "learning_rate": 5.665368421052632e-05,
-      "loss": 23.0,
-      "step": 93
-    },
-    {
-      "epoch": 0.003419549638037033,
-      "grad_norm": 0.0003987826348748058,
-      "learning_rate": 5.612421052631579e-05,
-      "loss": 23.0,
-      "step": 94
-    },
-    {
-      "epoch": 0.003455927825675725,
-      "grad_norm": 0.0005160235450603068,
-      "learning_rate": 5.559473684210527e-05,
-      "loss": 23.0,
-      "step": 95
-    },
-    {
-      "epoch": 0.0034923060133144166,
-      "grad_norm": 0.0004344135813880712,
-      "learning_rate": 5.506526315789474e-05,
-      "loss": 23.0,
-      "step": 96
-    },
-    {
-      "epoch": 0.0035286842009531086,
-      "grad_norm": 0.00038862781366333365,
-      "learning_rate": 5.453578947368421e-05,
-      "loss": 23.0,
-      "step": 97
-    },
-    {
-      "epoch": 0.0035650623885918,
-      "grad_norm": 0.0002816914056893438,
-      "learning_rate": 5.400631578947369e-05,
-      "loss": 23.0,
-      "step": 98
-    },
-    {
-      "epoch": 0.003601440576230492,
-      "grad_norm": 0.0007659445982426405,
-      "learning_rate": 5.347684210526316e-05,
-      "loss": 23.0,
-      "step": 99
-    },
-    {
-      "epoch": 0.003637818763869184,
-      "grad_norm": 0.0005744769005104899,
-      "learning_rate": 5.294736842105263e-05,
-      "loss": 23.0,
-      "step": 100
-    },
-    {
-      "epoch": 0.003637818763869184,
-      "eval_loss": 11.5,
-      "eval_runtime": 151.165,
-      "eval_samples_per_second": 76.572,
-      "eval_steps_per_second": 19.145,
-      "step": 100
-    },
-    {
-      "epoch": 0.0036741969515078758,
-      "grad_norm": 7.865398220019415e-05,
-      "learning_rate": 5.24178947368421e-05,
-      "loss": 23.0,
-      "step": 101
-    },
-    {
-      "epoch": 0.0037105751391465678,
-      "grad_norm": 0.00019760217401199043,
-      "learning_rate": 5.1888421052631585e-05,
-      "loss": 23.0,
-      "step": 102
-    },
-    {
-      "epoch": 0.0037469533267852594,
-      "grad_norm": 0.00019329931819811463,
-      "learning_rate": 5.135894736842106e-05,
-      "loss": 23.0,
-      "step": 103
-    },
-    {
-      "epoch": 0.0037833315144239514,
-      "grad_norm": 0.00010216799273621291,
-      "learning_rate": 5.082947368421053e-05,
-      "loss": 23.0,
-      "step": 104
-    },
-    {
-      "epoch": 0.0038197097020626434,
-      "grad_norm": 0.0001492125156801194,
-      "learning_rate": 5.03e-05,
-      "loss": 23.0,
-      "step": 105
-    },
-    {
-      "epoch": 0.003856087889701335,
-      "grad_norm": 0.000129396838019602,
-      "learning_rate": 4.977052631578947e-05,
-      "loss": 23.0,
-      "step": 106
-    },
-    {
-      "epoch": 0.003892466077340027,
-      "grad_norm": 0.00021847736206836998,
-      "learning_rate": 4.924105263157895e-05,
-      "loss": 23.0,
-      "step": 107
-    },
-    {
-      "epoch": 0.003928844264978719,
-      "grad_norm": 0.00036707802792079747,
-      "learning_rate": 4.871157894736843e-05,
-      "loss": 23.0,
-      "step": 108
-    },
-    {
-      "epoch": 0.003965222452617411,
-      "grad_norm": 0.00036315497709438205,
-      "learning_rate": 4.81821052631579e-05,
-      "loss": 23.0,
-      "step": 109
-    },
-    {
-      "epoch": 0.004001600640256103,
-      "grad_norm": 0.0002984280581586063,
-      "learning_rate": 4.765263157894737e-05,
-      "loss": 23.0,
-      "step": 110
-    },
-    {
-      "epoch": 0.004037978827894795,
-      "grad_norm": 0.0002477980451658368,
-      "learning_rate": 4.7123157894736845e-05,
-      "loss": 23.0,
-      "step": 111
-    },
-    {
-      "epoch": 0.004074357015533486,
-      "grad_norm": 0.00043583058868534863,
-      "learning_rate": 4.6593684210526316e-05,
-      "loss": 23.0,
-      "step": 112
-    },
-    {
-      "epoch": 0.004110735203172178,
-      "grad_norm": 0.00018492768867872655,
-      "learning_rate": 4.606421052631579e-05,
-      "loss": 23.0,
-      "step": 113
-    },
-    {
-      "epoch": 0.00414711339081087,
-      "grad_norm": 0.0003428026975598186,
-      "learning_rate": 4.553473684210527e-05,
-      "loss": 23.0,
-      "step": 114
-    },
-    {
-      "epoch": 0.004183491578449562,
-      "grad_norm": 9.029536886373535e-05,
-      "learning_rate": 4.500526315789474e-05,
-      "loss": 23.0,
-      "step": 115
-    },
-    {
-      "epoch": 0.004219869766088254,
-      "grad_norm": 0.00022797966084908694,
-      "learning_rate": 4.447578947368421e-05,
-      "loss": 23.0,
-      "step": 116
-    },
-    {
-      "epoch": 0.004256247953726945,
-      "grad_norm": 0.00019539303320925683,
-      "learning_rate": 4.394631578947369e-05,
-      "loss": 23.0,
-      "step": 117
-    },
-    {
-      "epoch": 0.004292626141365637,
-      "grad_norm": 0.00042453594505786896,
-      "learning_rate": 4.341684210526316e-05,
-      "loss": 23.0,
-      "step": 118
-    },
-    {
-      "epoch": 0.004329004329004329,
-      "grad_norm": 0.0002705013903323561,
-      "learning_rate": 4.2887368421052636e-05,
-      "loss": 23.0,
-      "step": 119
-    },
-    {
-      "epoch": 0.004365382516643021,
-      "grad_norm": 0.00021338310034479946,
-      "learning_rate": 4.2357894736842106e-05,
-      "loss": 23.0,
-      "step": 120
-    },
-    {
-      "epoch": 0.004401760704281713,
-      "grad_norm": 0.0002350714203203097,
-      "learning_rate": 4.182842105263158e-05,
-      "loss": 23.0,
-      "step": 121
-    },
-    {
-      "epoch": 0.004438138891920404,
-      "grad_norm": 0.0005433229962363839,
-      "learning_rate": 4.1298947368421053e-05,
-      "loss": 23.0,
-      "step": 122
-    },
-    {
-      "epoch": 0.004474517079559096,
-      "grad_norm": 0.0003798740799538791,
-      "learning_rate": 4.0769473684210524e-05,
-      "loss": 23.0,
-      "step": 123
-    },
-    {
-      "epoch": 0.004510895267197788,
-      "grad_norm": 0.0006719142547808588,
-      "learning_rate": 4.024000000000001e-05,
-      "loss": 23.0,
-      "step": 124
-    },
-    {
-      "epoch": 0.00454727345483648,
-      "grad_norm": 0.0005423346301540732,
-      "learning_rate": 3.971052631578948e-05,
-      "loss": 23.0,
-      "step": 125
-    },
-    {
-      "epoch": 0.004583651642475172,
-      "grad_norm": 0.0005073483916930854,
-      "learning_rate": 3.918105263157895e-05,
-      "loss": 23.0,
-      "step": 126
-    },
-    {
-      "epoch": 0.004620029830113863,
-      "grad_norm": 0.0005138374399393797,
-      "learning_rate": 3.865157894736842e-05,
-      "loss": 23.0,
-      "step": 127
-    },
-    {
-      "epoch": 0.004656408017752555,
-      "grad_norm": 0.0006768506718799472,
-      "learning_rate": 3.8122105263157896e-05,
-      "loss": 23.0,
-      "step": 128
-    },
-    {
-      "epoch": 0.004692786205391247,
-      "grad_norm": 0.0005319022457115352,
-      "learning_rate": 3.759263157894737e-05,
-      "loss": 23.0,
-      "step": 129
-    },
-    {
-      "epoch": 0.0047291643930299394,
-      "grad_norm": 0.0002845210547093302,
-      "learning_rate": 3.7063157894736844e-05,
-      "loss": 23.0,
-      "step": 130
-    },
-    {
-      "epoch": 0.0047655425806686314,
-      "grad_norm": 0.0006470263469964266,
-      "learning_rate": 3.653368421052632e-05,
-      "loss": 23.0,
-      "step": 131
-    },
-    {
-      "epoch": 0.004801920768307323,
-      "grad_norm": 0.00025029602693393826,
-      "learning_rate": 3.600421052631579e-05,
-      "loss": 23.0,
-      "step": 132
-    },
-    {
-      "epoch": 0.004838298955946015,
-      "grad_norm": 0.0005236798897385597,
-      "learning_rate": 3.547473684210526e-05,
-      "loss": 23.0,
-      "step": 133
-    },
-    {
-      "epoch": 0.004874677143584707,
-      "grad_norm": 0.00038987890002317727,
-      "learning_rate": 3.494526315789474e-05,
-      "loss": 23.0,
-      "step": 134
-    },
-    {
-      "epoch": 0.004911055331223399,
-      "grad_norm": 0.00047775980783626437,
-      "learning_rate": 3.4415789473684216e-05,
-      "loss": 23.0,
-      "step": 135
-    },
-    {
-      "epoch": 0.004947433518862091,
-      "grad_norm": 0.0008346535032615066,
-      "learning_rate": 3.3886315789473686e-05,
-      "loss": 23.0,
-      "step": 136
-    },
-    {
-      "epoch": 0.004983811706500782,
-      "grad_norm": 0.0005931820487603545,
-      "learning_rate": 3.3356842105263156e-05,
-      "loss": 23.0,
-      "step": 137
-    },
-    {
-      "epoch": 0.005020189894139474,
-      "grad_norm": 0.0006308462470769882,
-      "learning_rate": 3.2827368421052634e-05,
-      "loss": 23.0,
-      "step": 138
-    },
-    {
-      "epoch": 0.005056568081778166,
-      "grad_norm": 0.00030234106816351414,
-      "learning_rate": 3.2297894736842104e-05,
-      "loss": 23.0,
-      "step": 139
-    },
-    {
-      "epoch": 0.005092946269416858,
-      "grad_norm": 0.00049857038538903,
-      "learning_rate": 3.176842105263158e-05,
-      "loss": 23.0,
-      "step": 140
-    },
-    {
-      "epoch": 0.00512932445705555,
-      "grad_norm": 0.0003955921274609864,
-      "learning_rate": 3.123894736842106e-05,
-      "loss": 23.0,
-      "step": 141
-    },
-    {
-      "epoch": 0.005165702644694241,
-      "grad_norm": 0.000981033779680729,
-      "learning_rate": 3.070947368421053e-05,
-      "loss": 23.0,
-      "step": 142
-    },
-    {
-      "epoch": 0.005202080832332933,
-      "grad_norm": 0.0009714752668514848,
-      "learning_rate": 3.018e-05,
-      "loss": 23.0,
-      "step": 143
-    },
-    {
-      "epoch": 0.005238459019971625,
-      "grad_norm": 0.0006258689099922776,
-      "learning_rate": 2.9650526315789473e-05,
-      "loss": 23.0,
-      "step": 144
-    },
-    {
-      "epoch": 0.005274837207610317,
-      "grad_norm": 0.0005813623429276049,
-      "learning_rate": 2.912105263157895e-05,
-      "loss": 23.0,
-      "step": 145
-    },
-    {
-      "epoch": 0.005311215395249009,
-      "grad_norm": 0.0007660622941330075,
-      "learning_rate": 2.8591578947368424e-05,
-      "loss": 23.0,
-      "step": 146
-    },
-    {
-      "epoch": 0.0053475935828877,
-      "grad_norm": 0.0005416110507212579,
-      "learning_rate": 2.8062105263157894e-05,
-      "loss": 23.0,
-      "step": 147
-    },
-    {
-      "epoch": 0.005383971770526392,
-      "grad_norm": 0.0005930187180638313,
-      "learning_rate": 2.753263157894737e-05,
-      "loss": 23.0,
-      "step": 148
-    },
-    {
-      "epoch": 0.005420349958165084,
-      "grad_norm": 0.00032056248164735734,
-      "learning_rate": 2.7003157894736845e-05,
-      "loss": 23.0,
-      "step": 149
-    },
-    {
-      "epoch": 0.005456728145803776,
-      "grad_norm": 0.00021866206952836365,
-      "learning_rate": 2.6473684210526315e-05,
-      "loss": 23.0,
-      "step": 150
-    },
-    {
-      "epoch": 0.005456728145803776,
-      "eval_loss": 11.5,
-      "eval_runtime": 151.2013,
-      "eval_samples_per_second": 76.554,
-      "eval_steps_per_second": 19.14,
-      "step": 150
-    },
-    {
-      "epoch": 0.005493106333442468,
-      "grad_norm": 0.00036631166585721076,
-      "learning_rate": 2.5944210526315793e-05,
-      "loss": 23.0,
-      "step": 151
-    },
-    {
-      "epoch": 0.005529484521081159,
-      "grad_norm": 0.00015086479834280908,
-      "learning_rate": 2.5414736842105266e-05,
-      "loss": 23.0,
-      "step": 152
-    },
-    {
-      "epoch": 0.0055658627087198514,
-      "grad_norm": 0.00014822417870163918,
-      "learning_rate": 2.4885263157894737e-05,
-      "loss": 23.0,
-      "step": 153
-    },
-    {
-      "epoch": 0.0056022408963585435,
-      "grad_norm": 0.0002642092586029321,
-      "learning_rate": 2.4355789473684214e-05,
-      "loss": 23.0,
-      "step": 154
-    },
-    {
-      "epoch": 0.0056386190839972355,
-      "grad_norm": 0.0001480352075304836,
-      "learning_rate": 2.3826315789473684e-05,
-      "loss": 23.0,
-      "step": 155
-    },
-    {
-      "epoch": 0.0056749972716359275,
-      "grad_norm": 0.0002700032200664282,
-      "learning_rate": 2.3296842105263158e-05,
-      "loss": 23.0,
-      "step": 156
-    },
-    {
-      "epoch": 0.005711375459274619,
-      "grad_norm": 0.00038544239941984415,
-      "learning_rate": 2.2767368421052635e-05,
-      "loss": 23.0,
-      "step": 157
-    },
-    {
-      "epoch": 0.005747753646913311,
-      "grad_norm": 0.0004912279546260834,
-      "learning_rate": 2.2237894736842105e-05,
-      "loss": 23.0,
-      "step": 158
-    },
-    {
-      "epoch": 0.005784131834552003,
-      "grad_norm": 0.00031220720848068595,
-      "learning_rate": 2.170842105263158e-05,
-      "loss": 23.0,
-      "step": 159
-    },
-    {
-      "epoch": 0.005820510022190695,
-      "grad_norm": 0.0002950205816887319,
-      "learning_rate": 2.1178947368421053e-05,
-      "loss": 23.0,
-      "step": 160
-    },
-    {
-      "epoch": 0.005856888209829387,
-      "grad_norm": 0.00036991629167459905,
-      "learning_rate": 2.0649473684210527e-05,
-      "loss": 23.0,
-      "step": 161
-    },
-    {
-      "epoch": 0.005893266397468078,
-      "grad_norm": 0.0002873014018405229,
-      "learning_rate": 2.0120000000000004e-05,
-      "loss": 23.0,
-      "step": 162
-    },
-    {
-      "epoch": 0.00592964458510677,
-      "grad_norm": 0.0011097885435447097,
-      "learning_rate": 1.9590526315789474e-05,
-      "loss": 23.0,
-      "step": 163
-    },
-    {
-      "epoch": 0.005966022772745462,
-      "grad_norm": 0.0004760735318996012,
-      "learning_rate": 1.9061052631578948e-05,
-      "loss": 23.0,
-      "step": 164
-    },
-    {
-      "epoch": 0.006002400960384154,
-      "grad_norm": 0.0005161721492186189,
-      "learning_rate": 1.8531578947368422e-05,
-      "loss": 23.0,
-      "step": 165
-    },
-    {
-      "epoch": 0.006038779148022846,
-      "grad_norm": 0.00017680494056548923,
-      "learning_rate": 1.8002105263157896e-05,
-      "loss": 23.0,
-      "step": 166
-    },
-    {
-      "epoch": 0.006075157335661537,
-      "grad_norm": 0.0006741180550307035,
-      "learning_rate": 1.747263157894737e-05,
-      "loss": 23.0,
-      "step": 167
-    },
-    {
-      "epoch": 0.006111535523300229,
-      "grad_norm": 0.00031311047496274114,
-      "learning_rate": 1.6943157894736843e-05,
-      "loss": 23.0,
-      "step": 168
-    },
-    {
-      "epoch": 0.006147913710938921,
-      "grad_norm": 0.0004291492805350572,
-      "learning_rate": 1.6413684210526317e-05,
-      "loss": 23.0,
-      "step": 169
-    },
-    {
-      "epoch": 0.006184291898577613,
-      "grad_norm": 0.0002491539344191551,
-      "learning_rate": 1.588421052631579e-05,
-      "loss": 23.0,
-      "step": 170
-    },
-    {
-      "epoch": 0.006220670086216305,
-      "grad_norm": 0.000621207058429718,
-      "learning_rate": 1.5354736842105264e-05,
-      "loss": 23.0,
-      "step": 171
-    },
-    {
-      "epoch": 0.006257048273854996,
-      "grad_norm": 0.0002547437616158277,
-      "learning_rate": 1.4825263157894736e-05,
-      "loss": 23.0,
-      "step": 172
-    },
-    {
-      "epoch": 0.006293426461493688,
-      "grad_norm": 0.0006873645470477641,
-      "learning_rate": 1.4295789473684212e-05,
-      "loss": 23.0,
-      "step": 173
-    },
-    {
-      "epoch": 0.00632980464913238,
-      "grad_norm": 0.0002644654014147818,
-      "learning_rate": 1.3766315789473686e-05,
-      "loss": 23.0,
-      "step": 174
-    },
-    {
-      "epoch": 0.006366182836771072,
-      "grad_norm": 0.0005771035212092102,
-      "learning_rate": 1.3236842105263158e-05,
-      "loss": 23.0,
-      "step": 175
-    },
-    {
-      "epoch": 0.006402561024409764,
-      "grad_norm": 0.000470922386739403,
-      "learning_rate": 1.2707368421052633e-05,
-      "loss": 23.0,
-      "step": 176
-    },
-    {
-      "epoch": 0.0064389392120484555,
-      "grad_norm": 0.0005482633132487535,
-      "learning_rate": 1.2177894736842107e-05,
-      "loss": 23.0,
-      "step": 177
-    },
-    {
-      "epoch": 0.0064753173996871475,
-      "grad_norm": 0.0002871015458367765,
-      "learning_rate": 1.1648421052631579e-05,
-      "loss": 23.0,
-      "step": 178
-    },
-    {
-      "epoch": 0.0065116955873258395,
-      "grad_norm": 0.0006182437646202743,
-      "learning_rate": 1.1118947368421053e-05,
-      "loss": 23.0,
-      "step": 179
-    },
-    {
-      "epoch": 0.0065480737749645315,
-      "grad_norm": 0.0006852405495010316,
-      "learning_rate": 1.0589473684210526e-05,
-      "loss": 23.0,
-      "step": 180
-    },
-    {
-      "epoch": 0.0065844519626032235,
-      "grad_norm": 0.0006683968240395188,
-      "learning_rate": 1.0060000000000002e-05,
-      "loss": 23.0,
-      "step": 181
-    },
-    {
-      "epoch": 0.006620830150241915,
-      "grad_norm": 0.0009705866104923189,
-      "learning_rate": 9.530526315789474e-06,
-      "loss": 23.0,
-      "step": 182
-    },
-    {
-      "epoch": 0.006657208337880607,
-      "grad_norm": 0.0004722289741039276,
-      "learning_rate": 9.001052631578948e-06,
-      "loss": 23.0,
-      "step": 183
-    },
-    {
-      "epoch": 0.006693586525519299,
-      "grad_norm": 0.0003802978608291596,
-      "learning_rate": 8.471578947368422e-06,
-      "loss": 23.0,
-      "step": 184
-    },
-    {
-      "epoch": 0.006729964713157991,
-      "grad_norm": 0.001799766207113862,
-      "learning_rate": 7.942105263157895e-06,
-      "loss": 23.0,
-      "step": 185
-    },
-    {
-      "epoch": 0.006766342900796682,
-      "grad_norm": 0.0008337293402291834,
-      "learning_rate": 7.412631578947368e-06,
-      "loss": 23.0,
-      "step": 186
-    },
-    {
-      "epoch": 0.006802721088435374,
-      "grad_norm": 0.0007884498336352408,
-      "learning_rate": 6.883157894736843e-06,
-      "loss": 23.0,
-      "step": 187
-    },
-    {
-      "epoch": 0.006839099276074066,
-      "grad_norm": 0.0008693984127603471,
-      "learning_rate": 6.3536842105263166e-06,
-      "loss": 23.0,
-      "step": 188
-    },
-    {
-      "epoch": 0.006875477463712758,
-      "grad_norm": 0.0009469325887039304,
-      "learning_rate": 5.8242105263157895e-06,
-      "loss": 23.0,
-      "step": 189
-    },
-    {
-      "epoch": 0.00691185565135145,
-      "grad_norm": 0.0007177354418672621,
-      "learning_rate": 5.294736842105263e-06,
-      "loss": 23.0,
-      "step": 190
-    },
-    {
-      "epoch": 0.006948233838990141,
-      "grad_norm": 0.0008088785689324141,
-      "learning_rate": 4.765263157894737e-06,
-      "loss": 23.0,
-      "step": 191
-    },
-    {
-      "epoch": 0.006984612026628833,
-      "grad_norm": 0.00048786430852487683,
-      "learning_rate": 4.235789473684211e-06,
-      "loss": 23.0,
-      "step": 192
-    },
-    {
-      "epoch": 0.007020990214267525,
-      "grad_norm": 0.00033468171022832394,
-      "learning_rate": 3.706315789473684e-06,
-      "loss": 23.0,
-      "step": 193
-    },
-    {
-      "epoch": 0.007057368401906217,
-      "grad_norm": 0.0012459014542400837,
-      "learning_rate": 3.1768421052631583e-06,
-      "loss": 23.0,
-      "step": 194
-    },
-    {
-      "epoch": 0.007093746589544909,
-      "grad_norm": 0.0007038541371002793,
-      "learning_rate": 2.6473684210526316e-06,
-      "loss": 23.0,
-      "step": 195
-    },
-    {
-      "epoch": 0.0071301247771836,
-      "grad_norm": 0.001085408846847713,
-      "learning_rate": 2.1178947368421054e-06,
-      "loss": 23.0,
-      "step": 196
-    },
-    {
-      "epoch": 0.007166502964822292,
-      "grad_norm": 0.0005441877874545753,
-      "learning_rate": 1.5884210526315791e-06,
-      "loss": 23.0,
-      "step": 197
-    },
-    {
-      "epoch": 0.007202881152460984,
-      "grad_norm": 0.0011435911292210221,
-      "learning_rate": 1.0589473684210527e-06,
-      "loss": 23.0,
-      "step": 198
-    },
-    {
-      "epoch": 0.007239259340099676,
-      "grad_norm": 0.000514986168127507,
-      "learning_rate": 5.294736842105263e-07,
-      "loss": 23.0,
-      "step": 199
-    },
-    {
-      "epoch": 0.007275637527738368,
-      "grad_norm": 0.0006262768292799592,
-      "learning_rate": 0.0,
-      "loss": 23.0,
-      "step": 200
-    },
-    {
-      "epoch": 0.007275637527738368,
-      "eval_loss": 11.5,
-      "eval_runtime": 204.8225,
-      "eval_samples_per_second": 56.512,
-      "eval_steps_per_second": 14.129,
-      "step": 200
     }
   ],
   "logging_steps": 1,
@@ -1461,7 +387,7 @@
         "early_stopping_threshold": 0.0
       },
       "attributes": {
-        "early_stopping_patience_counter": 3
+        "early_stopping_patience_counter": 0
       }
     },
     "TrainerControl": {
@@ -1470,12 +396,12 @@
         "should_evaluate": false,
         "should_log": false,
         "should_save": true,
-        "should_training_stop": true
+        "should_training_stop": false
       },
       "attributes": {}
     }
   },
-  "total_flos": 8059714732032.0,
+  "total_flos": 2030002962432.0,
   "train_batch_size": 4,
   "trial_name": null,
   "trial_params": null

last-checkpoint/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2f70492f7bfc0b5a2028ba0ef950958a67d03aee8a68c11c59c8c04032c230aa
+oid sha256:eff417ff5f179436b0b27b1f383c5d1106ad333a2a1841023fe6defd53aa6431
 size 6840