| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.9967914438502674, | |
| "eval_steps": 500, | |
| "global_step": 233, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "completion_length": 230.126953125, | |
| "epoch": 0.017112299465240642, | |
| "grad_norm": 0.4056285321712494, | |
| "kl": 0.0079345703125, | |
| "learning_rate": 9.82832618025751e-07, | |
| "loss": 0.0004, | |
| "reward": 0.015625, | |
| "reward_std": 0.03125, | |
| "rewards/validate_answer_with_correct_format": 0.005859375, | |
| "rewards/validate_format": 0.009765625, | |
| "step": 4 | |
| }, | |
| { | |
| "completion_length": 219.341796875, | |
| "epoch": 0.034224598930481284, | |
| "grad_norm": 0.26790499687194824, | |
| "kl": 0.016498565673828125, | |
| "learning_rate": 9.656652360515022e-07, | |
| "loss": 0.0008, | |
| "reward": 0.01171875, | |
| "reward_std": 0.0234375, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.0078125, | |
| "step": 8 | |
| }, | |
| { | |
| "completion_length": 221.603515625, | |
| "epoch": 0.051336898395721926, | |
| "grad_norm": 0.14115919172763824, | |
| "kl": 0.017139434814453125, | |
| "learning_rate": 9.484978540772532e-07, | |
| "loss": 0.0009, | |
| "reward": 0.005859375, | |
| "reward_std": 0.01171875, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 12 | |
| }, | |
| { | |
| "completion_length": 226.123046875, | |
| "epoch": 0.06844919786096257, | |
| "grad_norm": 0.8012773394584656, | |
| "kl": 0.028614044189453125, | |
| "learning_rate": 9.313304721030042e-07, | |
| "loss": 0.0014, | |
| "reward": 0.013671875, | |
| "reward_std": 0.02734375, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.009765625, | |
| "step": 16 | |
| }, | |
| { | |
| "completion_length": 219.775390625, | |
| "epoch": 0.0855614973262032, | |
| "grad_norm": 0.15161409974098206, | |
| "kl": 0.017520904541015625, | |
| "learning_rate": 9.141630901287554e-07, | |
| "loss": 0.0009, | |
| "reward": 0.01171875, | |
| "reward_std": 0.0234375, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.009765625, | |
| "step": 20 | |
| }, | |
| { | |
| "completion_length": 224.556640625, | |
| "epoch": 0.10267379679144385, | |
| "grad_norm": 0.13644962012767792, | |
| "kl": 0.0185394287109375, | |
| "learning_rate": 8.969957081545064e-07, | |
| "loss": 0.0009, | |
| "reward": 0.009765625, | |
| "reward_std": 0.01953125, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.0078125, | |
| "step": 24 | |
| }, | |
| { | |
| "completion_length": 223.056640625, | |
| "epoch": 0.11978609625668449, | |
| "grad_norm": 0.15331625938415527, | |
| "kl": 0.01702117919921875, | |
| "learning_rate": 8.798283261802575e-07, | |
| "loss": 0.0009, | |
| "reward": 0.001953125, | |
| "reward_std": 0.00390625, | |
| "rewards/validate_answer_with_correct_format": 0.0, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 28 | |
| }, | |
| { | |
| "completion_length": 225.7109375, | |
| "epoch": 0.13689839572192514, | |
| "grad_norm": 0.20857754349708557, | |
| "kl": 0.015987396240234375, | |
| "learning_rate": 8.626609442060086e-07, | |
| "loss": 0.0008, | |
| "reward": 0.015625, | |
| "reward_std": 0.027948048897087574, | |
| "rewards/validate_answer_with_correct_format": 0.0078125, | |
| "rewards/validate_format": 0.0078125, | |
| "step": 32 | |
| }, | |
| { | |
| "completion_length": 224.224609375, | |
| "epoch": 0.15401069518716579, | |
| "grad_norm": 0.2771243453025818, | |
| "kl": 0.017055511474609375, | |
| "learning_rate": 8.454935622317596e-07, | |
| "loss": 0.0009, | |
| "reward": 0.00390625, | |
| "reward_std": 0.0078125, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 36 | |
| }, | |
| { | |
| "completion_length": 222.896484375, | |
| "epoch": 0.1711229946524064, | |
| "grad_norm": 0.31459948420524597, | |
| "kl": 0.015277862548828125, | |
| "learning_rate": 8.283261802575107e-07, | |
| "loss": 0.0008, | |
| "reward": 0.0078125, | |
| "reward_std": 0.015625, | |
| "rewards/validate_answer_with_correct_format": 0.005859375, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 40 | |
| }, | |
| { | |
| "completion_length": 222.875, | |
| "epoch": 0.18823529411764706, | |
| "grad_norm": 0.7412174344062805, | |
| "kl": 0.01706695556640625, | |
| "learning_rate": 8.111587982832617e-07, | |
| "loss": 0.0009, | |
| "reward": 0.009765625, | |
| "reward_std": 0.01953125, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.005859375, | |
| "step": 44 | |
| }, | |
| { | |
| "completion_length": 224.634765625, | |
| "epoch": 0.2053475935828877, | |
| "grad_norm": 0.1542515754699707, | |
| "kl": 0.015285491943359375, | |
| "learning_rate": 7.939914163090128e-07, | |
| "loss": 0.0008, | |
| "reward": 0.009765625, | |
| "reward_std": 0.01953125, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.0078125, | |
| "step": 48 | |
| }, | |
| { | |
| "completion_length": 224.271484375, | |
| "epoch": 0.22245989304812835, | |
| "grad_norm": 0.2658785879611969, | |
| "kl": 0.014251708984375, | |
| "learning_rate": 7.76824034334764e-07, | |
| "loss": 0.0007, | |
| "reward": 0.00390625, | |
| "reward_std": 0.0078125, | |
| "rewards/validate_answer_with_correct_format": 0.0, | |
| "rewards/validate_format": 0.00390625, | |
| "step": 52 | |
| }, | |
| { | |
| "completion_length": 226.400390625, | |
| "epoch": 0.23957219251336898, | |
| "grad_norm": 0.08715300261974335, | |
| "kl": 0.01432037353515625, | |
| "learning_rate": 7.59656652360515e-07, | |
| "loss": 0.0007, | |
| "reward": 0.001953125, | |
| "reward_std": 0.00390625, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.0, | |
| "step": 56 | |
| }, | |
| { | |
| "completion_length": 227.474609375, | |
| "epoch": 0.25668449197860965, | |
| "grad_norm": 0.21552450954914093, | |
| "kl": 0.014652252197265625, | |
| "learning_rate": 7.424892703862661e-07, | |
| "loss": 0.0007, | |
| "reward": 0.0078125, | |
| "reward_std": 0.015625, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.005859375, | |
| "step": 60 | |
| }, | |
| { | |
| "completion_length": 226.896484375, | |
| "epoch": 0.2737967914438503, | |
| "grad_norm": 1.008454442024231, | |
| "kl": 0.014369964599609375, | |
| "learning_rate": 7.253218884120171e-07, | |
| "loss": 0.0007, | |
| "reward": 0.00390625, | |
| "reward_std": 0.0078125, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.0, | |
| "step": 64 | |
| }, | |
| { | |
| "completion_length": 224.822265625, | |
| "epoch": 0.2909090909090909, | |
| "grad_norm": 1.7959470748901367, | |
| "kl": 0.01448822021484375, | |
| "learning_rate": 7.081545064377682e-07, | |
| "loss": 0.0007, | |
| "reward": 0.0078125, | |
| "reward_std": 0.015625, | |
| "rewards/validate_answer_with_correct_format": 0.0, | |
| "rewards/validate_format": 0.0078125, | |
| "step": 68 | |
| }, | |
| { | |
| "completion_length": 222.81640625, | |
| "epoch": 0.30802139037433157, | |
| "grad_norm": 0.14670810103416443, | |
| "kl": 0.0145416259765625, | |
| "learning_rate": 6.909871244635192e-07, | |
| "loss": 0.0007, | |
| "reward": 0.005859375, | |
| "reward_std": 0.01171875, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 72 | |
| }, | |
| { | |
| "completion_length": 223.57421875, | |
| "epoch": 0.3251336898395722, | |
| "grad_norm": 0.2493465095758438, | |
| "kl": 0.01750946044921875, | |
| "learning_rate": 6.738197424892703e-07, | |
| "loss": 0.0009, | |
| "reward": 0.009765625, | |
| "reward_std": 0.01953125, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.005859375, | |
| "step": 76 | |
| }, | |
| { | |
| "completion_length": 223.986328125, | |
| "epoch": 0.3422459893048128, | |
| "grad_norm": 0.09515038132667542, | |
| "kl": 0.01554107666015625, | |
| "learning_rate": 6.566523605150214e-07, | |
| "loss": 0.0008, | |
| "reward": 0.00390625, | |
| "reward_std": 0.0078125, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 80 | |
| }, | |
| { | |
| "completion_length": 226.271484375, | |
| "epoch": 0.3593582887700535, | |
| "grad_norm": 0.3518137037754059, | |
| "kl": 0.01836395263671875, | |
| "learning_rate": 6.394849785407725e-07, | |
| "loss": 0.0009, | |
| "reward": 0.00390625, | |
| "reward_std": 0.0078125, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 84 | |
| }, | |
| { | |
| "completion_length": 222.298828125, | |
| "epoch": 0.3764705882352941, | |
| "grad_norm": 0.22661465406417847, | |
| "kl": 0.015354156494140625, | |
| "learning_rate": 6.223175965665236e-07, | |
| "loss": 0.0008, | |
| "reward": 0.013671875, | |
| "reward_std": 0.02734375, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.01171875, | |
| "step": 88 | |
| }, | |
| { | |
| "completion_length": 224.556640625, | |
| "epoch": 0.39358288770053473, | |
| "grad_norm": 0.530048668384552, | |
| "kl": 0.014934539794921875, | |
| "learning_rate": 6.051502145922746e-07, | |
| "loss": 0.0007, | |
| "reward": 0.01171875, | |
| "reward_std": 0.0234375, | |
| "rewards/validate_answer_with_correct_format": 0.0078125, | |
| "rewards/validate_format": 0.00390625, | |
| "step": 92 | |
| }, | |
| { | |
| "completion_length": 224.291015625, | |
| "epoch": 0.4106951871657754, | |
| "grad_norm": 0.5434131026268005, | |
| "kl": 0.01446533203125, | |
| "learning_rate": 5.879828326180257e-07, | |
| "loss": 0.0007, | |
| "reward": 0.009765625, | |
| "reward_std": 0.01953125, | |
| "rewards/validate_answer_with_correct_format": 0.005859375, | |
| "rewards/validate_format": 0.00390625, | |
| "step": 96 | |
| }, | |
| { | |
| "completion_length": 221.537109375, | |
| "epoch": 0.42780748663101603, | |
| "grad_norm": 1.3832066059112549, | |
| "kl": 0.017185211181640625, | |
| "learning_rate": 5.708154506437767e-07, | |
| "loss": 0.0009, | |
| "reward": 0.0078125, | |
| "reward_std": 0.015625, | |
| "rewards/validate_answer_with_correct_format": 0.005859375, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 100 | |
| }, | |
| { | |
| "completion_length": 224.041015625, | |
| "epoch": 0.4449197860962567, | |
| "grad_norm": 0.299678236246109, | |
| "kl": 0.019840240478515625, | |
| "learning_rate": 5.536480686695278e-07, | |
| "loss": 0.001, | |
| "reward": 0.0078125, | |
| "reward_std": 0.015625, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.00390625, | |
| "step": 104 | |
| }, | |
| { | |
| "completion_length": 222.169921875, | |
| "epoch": 0.46203208556149733, | |
| "grad_norm": 1.4706224203109741, | |
| "kl": 0.018634796142578125, | |
| "learning_rate": 5.364806866952789e-07, | |
| "loss": 0.0009, | |
| "reward": 0.009765625, | |
| "reward_std": 0.01953125, | |
| "rewards/validate_answer_with_correct_format": 0.005859375, | |
| "rewards/validate_format": 0.00390625, | |
| "step": 108 | |
| }, | |
| { | |
| "completion_length": 220.880859375, | |
| "epoch": 0.47914438502673795, | |
| "grad_norm": 0.49486151337623596, | |
| "kl": 0.019145965576171875, | |
| "learning_rate": 5.193133047210299e-07, | |
| "loss": 0.001, | |
| "reward": 0.005859375, | |
| "reward_std": 0.01171875, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 112 | |
| }, | |
| { | |
| "completion_length": 228.208984375, | |
| "epoch": 0.49625668449197863, | |
| "grad_norm": 0.23438626527786255, | |
| "kl": 0.0195465087890625, | |
| "learning_rate": 5.021459227467812e-07, | |
| "loss": 0.001, | |
| "reward": 0.0078125, | |
| "reward_std": 0.015625, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.005859375, | |
| "step": 116 | |
| }, | |
| { | |
| "completion_length": 225.1484375, | |
| "epoch": 0.5133689839572193, | |
| "grad_norm": 0.17298594117164612, | |
| "kl": 0.017486572265625, | |
| "learning_rate": 4.849785407725322e-07, | |
| "loss": 0.0009, | |
| "reward": 0.009765625, | |
| "reward_std": 0.016229298897087574, | |
| "rewards/validate_answer_with_correct_format": 0.005859375, | |
| "rewards/validate_format": 0.00390625, | |
| "step": 120 | |
| }, | |
| { | |
| "completion_length": 226.361328125, | |
| "epoch": 0.5304812834224599, | |
| "grad_norm": 0.5948027968406677, | |
| "kl": 0.019260406494140625, | |
| "learning_rate": 4.6781115879828326e-07, | |
| "loss": 0.001, | |
| "reward": 0.005859375, | |
| "reward_std": 0.01171875, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.00390625, | |
| "step": 124 | |
| }, | |
| { | |
| "completion_length": 224.271484375, | |
| "epoch": 0.5475935828877005, | |
| "grad_norm": 0.44078490138053894, | |
| "kl": 0.020477294921875, | |
| "learning_rate": 4.506437768240343e-07, | |
| "loss": 0.001, | |
| "reward": 0.0078125, | |
| "reward_std": 0.012323048897087574, | |
| "rewards/validate_answer_with_correct_format": 0.0, | |
| "rewards/validate_format": 0.0078125, | |
| "step": 128 | |
| }, | |
| { | |
| "completion_length": 228.298828125, | |
| "epoch": 0.5647058823529412, | |
| "grad_norm": 0.2306816428899765, | |
| "kl": 0.061435699462890625, | |
| "learning_rate": 4.3347639484978536e-07, | |
| "loss": 0.0031, | |
| "reward": 0.015625, | |
| "reward_std": 0.03125, | |
| "rewards/validate_answer_with_correct_format": 0.0078125, | |
| "rewards/validate_format": 0.0078125, | |
| "step": 132 | |
| }, | |
| { | |
| "completion_length": 223.705078125, | |
| "epoch": 0.5818181818181818, | |
| "grad_norm": 0.4965137839317322, | |
| "kl": 0.0200042724609375, | |
| "learning_rate": 4.163090128755364e-07, | |
| "loss": 0.001, | |
| "reward": 0.017578125, | |
| "reward_std": 0.03515625, | |
| "rewards/validate_answer_with_correct_format": 0.009765625, | |
| "rewards/validate_format": 0.0078125, | |
| "step": 136 | |
| }, | |
| { | |
| "completion_length": 219.962890625, | |
| "epoch": 0.5989304812834224, | |
| "grad_norm": 0.13467183709144592, | |
| "kl": 0.019496917724609375, | |
| "learning_rate": 3.991416309012876e-07, | |
| "loss": 0.001, | |
| "reward": 0.00390625, | |
| "reward_std": 0.0078125, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.0, | |
| "step": 140 | |
| }, | |
| { | |
| "completion_length": 224.228515625, | |
| "epoch": 0.6160427807486631, | |
| "grad_norm": 0.7464412450790405, | |
| "kl": 0.019817352294921875, | |
| "learning_rate": 3.819742489270386e-07, | |
| "loss": 0.001, | |
| "reward": 0.009765625, | |
| "reward_std": 0.01953125, | |
| "rewards/validate_answer_with_correct_format": 0.005859375, | |
| "rewards/validate_format": 0.00390625, | |
| "step": 144 | |
| }, | |
| { | |
| "completion_length": 227.57421875, | |
| "epoch": 0.6331550802139038, | |
| "grad_norm": 0.2531176209449768, | |
| "kl": 0.02042388916015625, | |
| "learning_rate": 3.648068669527897e-07, | |
| "loss": 0.001, | |
| "reward": 0.001953125, | |
| "reward_std": 0.00390625, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.0, | |
| "step": 148 | |
| }, | |
| { | |
| "completion_length": 222.51171875, | |
| "epoch": 0.6502673796791444, | |
| "grad_norm": 0.11502356827259064, | |
| "kl": 0.01702117919921875, | |
| "learning_rate": 3.4763948497854073e-07, | |
| "loss": 0.0009, | |
| "reward": 0.00390625, | |
| "reward_std": 0.0078125, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 152 | |
| }, | |
| { | |
| "completion_length": 222.107421875, | |
| "epoch": 0.667379679144385, | |
| "grad_norm": 0.18971668183803558, | |
| "kl": 0.01854705810546875, | |
| "learning_rate": 3.3047210300429184e-07, | |
| "loss": 0.0009, | |
| "reward": 0.01171875, | |
| "reward_std": 0.0234375, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.0078125, | |
| "step": 156 | |
| }, | |
| { | |
| "completion_length": 224.759765625, | |
| "epoch": 0.6844919786096256, | |
| "grad_norm": 0.07911235839128494, | |
| "kl": 0.0183258056640625, | |
| "learning_rate": 3.133047210300429e-07, | |
| "loss": 0.0009, | |
| "reward": 0.0, | |
| "reward_std": 0.0, | |
| "rewards/validate_answer_with_correct_format": 0.0, | |
| "rewards/validate_format": 0.0, | |
| "step": 160 | |
| }, | |
| { | |
| "completion_length": 222.0546875, | |
| "epoch": 0.7016042780748664, | |
| "grad_norm": 0.957903265953064, | |
| "kl": 0.01689910888671875, | |
| "learning_rate": 2.96137339055794e-07, | |
| "loss": 0.0008, | |
| "reward": 0.0078125, | |
| "reward_std": 0.015625, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.00390625, | |
| "step": 164 | |
| }, | |
| { | |
| "completion_length": 221.697265625, | |
| "epoch": 0.718716577540107, | |
| "grad_norm": 0.13198330998420715, | |
| "kl": 1.3442916870117188, | |
| "learning_rate": 2.7896995708154505e-07, | |
| "loss": 0.0671, | |
| "reward": 0.0, | |
| "reward_std": 0.0, | |
| "rewards/validate_answer_with_correct_format": 0.0, | |
| "rewards/validate_format": 0.0, | |
| "step": 168 | |
| }, | |
| { | |
| "completion_length": 225.87890625, | |
| "epoch": 0.7358288770053476, | |
| "grad_norm": 0.2404724359512329, | |
| "kl": 0.019535064697265625, | |
| "learning_rate": 2.6180257510729615e-07, | |
| "loss": 0.001, | |
| "reward": 0.0078125, | |
| "reward_std": 0.015625, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.00390625, | |
| "step": 172 | |
| }, | |
| { | |
| "completion_length": 224.82421875, | |
| "epoch": 0.7529411764705882, | |
| "grad_norm": 0.07936124503612518, | |
| "kl": 0.01831817626953125, | |
| "learning_rate": 2.446351931330472e-07, | |
| "loss": 0.0009, | |
| "reward": 0.0078125, | |
| "reward_std": 0.012323048897087574, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.005859375, | |
| "step": 176 | |
| }, | |
| { | |
| "completion_length": 222.529296875, | |
| "epoch": 0.7700534759358288, | |
| "grad_norm": 0.23834188282489777, | |
| "kl": 0.018939971923828125, | |
| "learning_rate": 2.2746781115879825e-07, | |
| "loss": 0.0009, | |
| "reward": 0.0078125, | |
| "reward_std": 0.015625, | |
| "rewards/validate_answer_with_correct_format": 0.005859375, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 180 | |
| }, | |
| { | |
| "completion_length": 224.953125, | |
| "epoch": 0.7871657754010695, | |
| "grad_norm": 0.8199672698974609, | |
| "kl": 0.01871490478515625, | |
| "learning_rate": 2.1030042918454936e-07, | |
| "loss": 0.0009, | |
| "reward": 0.009765625, | |
| "reward_std": 0.01953125, | |
| "rewards/validate_answer_with_correct_format": 0.005859375, | |
| "rewards/validate_format": 0.00390625, | |
| "step": 184 | |
| }, | |
| { | |
| "completion_length": 228.21484375, | |
| "epoch": 0.8042780748663102, | |
| "grad_norm": 0.1074649915099144, | |
| "kl": 0.01839447021484375, | |
| "learning_rate": 1.931330472103004e-07, | |
| "loss": 0.0009, | |
| "reward": 0.001953125, | |
| "reward_std": 0.00390625, | |
| "rewards/validate_answer_with_correct_format": 0.0, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 188 | |
| }, | |
| { | |
| "completion_length": 228.591796875, | |
| "epoch": 0.8213903743315508, | |
| "grad_norm": 0.16989368200302124, | |
| "kl": 0.016857147216796875, | |
| "learning_rate": 1.759656652360515e-07, | |
| "loss": 0.0008, | |
| "reward": 0.00390625, | |
| "reward_std": 0.004510548897087574, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.0, | |
| "step": 192 | |
| }, | |
| { | |
| "completion_length": 224.1484375, | |
| "epoch": 0.8385026737967914, | |
| "grad_norm": 0.2016996294260025, | |
| "kl": 0.017910003662109375, | |
| "learning_rate": 1.5879828326180257e-07, | |
| "loss": 0.0009, | |
| "reward": 0.001953125, | |
| "reward_std": 0.00390625, | |
| "rewards/validate_answer_with_correct_format": 0.0, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 196 | |
| }, | |
| { | |
| "completion_length": 223.845703125, | |
| "epoch": 0.8556149732620321, | |
| "grad_norm": 0.09751327335834503, | |
| "kl": 0.0189056396484375, | |
| "learning_rate": 1.4163090128755365e-07, | |
| "loss": 0.0009, | |
| "reward": 0.00390625, | |
| "reward_std": 0.0078125, | |
| "rewards/validate_answer_with_correct_format": 0.001953125, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 200 | |
| }, | |
| { | |
| "completion_length": 225.376953125, | |
| "epoch": 0.8727272727272727, | |
| "grad_norm": 0.7625419497489929, | |
| "kl": 0.016994476318359375, | |
| "learning_rate": 1.2446351931330473e-07, | |
| "loss": 0.0008, | |
| "reward": 0.009765625, | |
| "reward_std": 0.01953125, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.005859375, | |
| "step": 204 | |
| }, | |
| { | |
| "completion_length": 222.322265625, | |
| "epoch": 0.8898395721925134, | |
| "grad_norm": 0.4644334018230438, | |
| "kl": 0.017730712890625, | |
| "learning_rate": 1.0729613733905579e-07, | |
| "loss": 0.0009, | |
| "reward": 0.0078125, | |
| "reward_std": 0.015625, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.00390625, | |
| "step": 208 | |
| }, | |
| { | |
| "completion_length": 227.466796875, | |
| "epoch": 0.906951871657754, | |
| "grad_norm": 0.36913207173347473, | |
| "kl": 0.0170745849609375, | |
| "learning_rate": 9.012875536480687e-08, | |
| "loss": 0.0009, | |
| "reward": 0.009765625, | |
| "reward_std": 0.01953125, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.005859375, | |
| "step": 212 | |
| }, | |
| { | |
| "completion_length": 229.888671875, | |
| "epoch": 0.9240641711229947, | |
| "grad_norm": 0.09142182767391205, | |
| "kl": 0.016002655029296875, | |
| "learning_rate": 7.296137339055794e-08, | |
| "loss": 0.0008, | |
| "reward": 0.0, | |
| "reward_std": 0.0, | |
| "rewards/validate_answer_with_correct_format": 0.0, | |
| "rewards/validate_format": 0.0, | |
| "step": 216 | |
| }, | |
| { | |
| "completion_length": 225.806640625, | |
| "epoch": 0.9411764705882353, | |
| "grad_norm": 0.4567970931529999, | |
| "kl": 0.01589202880859375, | |
| "learning_rate": 5.5793991416309014e-08, | |
| "loss": 0.0008, | |
| "reward": 0.0078125, | |
| "reward_std": 0.012323048897087574, | |
| "rewards/validate_answer_with_correct_format": 0.005859375, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 220 | |
| }, | |
| { | |
| "completion_length": 227.38671875, | |
| "epoch": 0.9582887700534759, | |
| "grad_norm": 0.08972982317209244, | |
| "kl": 0.017608642578125, | |
| "learning_rate": 3.8626609442060086e-08, | |
| "loss": 0.0009, | |
| "reward": 0.0, | |
| "reward_std": 0.0, | |
| "rewards/validate_answer_with_correct_format": 0.0, | |
| "rewards/validate_format": 0.0, | |
| "step": 224 | |
| }, | |
| { | |
| "completion_length": 224.208984375, | |
| "epoch": 0.9754010695187165, | |
| "grad_norm": 0.1748742312192917, | |
| "kl": 0.017276763916015625, | |
| "learning_rate": 2.1459227467811158e-08, | |
| "loss": 0.0009, | |
| "reward": 0.0078125, | |
| "reward_std": 0.015625, | |
| "rewards/validate_answer_with_correct_format": 0.00390625, | |
| "rewards/validate_format": 0.00390625, | |
| "step": 228 | |
| }, | |
| { | |
| "completion_length": 222.94921875, | |
| "epoch": 0.9925133689839573, | |
| "grad_norm": 0.1176988035440445, | |
| "kl": 0.01596832275390625, | |
| "learning_rate": 4.291845493562231e-09, | |
| "loss": 0.0008, | |
| "reward": 0.009765625, | |
| "reward_std": 0.01953125, | |
| "rewards/validate_answer_with_correct_format": 0.0078125, | |
| "rewards/validate_format": 0.001953125, | |
| "step": 232 | |
| } | |
| ], | |
| "logging_steps": 4, | |
| "max_steps": 233, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 1, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 0.0, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |