{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 0.5, "eval_steps": 50, "global_step": 234, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0201171875, "completions/max_length": 4096.0, "completions/max_terminated_length": 3988.6, "completions/mean_length": 583.637890625, "completions/mean_terminated_length": 511.51544799804685, "completions/min_length": 4.0, "completions/min_terminated_length": 4.0, "entropy": 0.2013093136018142, "epoch": 0.010683760683760684, "frac_reward_zero_std": 0.003125, "grad_norm": 0.37152326107025146, "learning_rate": 4.2553191489361704e-07, "loss": 0.0047, "num_tokens": 7689764.0, "reward": 0.7603515625, "reward_std": 0.8043357372283936, "rewards/accuracy_reward/mean": 0.23271484375, "rewards/accuracy_reward/std": 0.4221785068511963, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.52763671875, "rewards/format_reward/std": 0.49887575507164, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 5, "step_time": 322.47677695861785 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0212890625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4037.4, "completions/mean_length": 584.3814453125, "completions/mean_terminated_length": 508.05540161132814, "completions/min_length": 3.2, "completions/min_terminated_length": 3.2, "entropy": 0.2046258968883194, "epoch": 0.021367521367521368, "frac_reward_zero_std": 0.003125, "grad_norm": 8.443074226379395, "learning_rate": 9.574468085106384e-07, "loss": 0.0061, "num_tokens": 15371334.0, "reward": 0.7640625, "reward_std": 0.806778860092163, "rewards/accuracy_reward/mean": 0.23662109375, "rewards/accuracy_reward/std": 0.4241984665393829, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.52744140625, "rewards/format_reward/std": 0.49808618426322937, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 10, "step_time": 279.42059247897123 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0216796875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4012.0, "completions/mean_length": 584.8390625, "completions/mean_terminated_length": 507.1805419921875, "completions/min_length": 3.2, "completions/min_terminated_length": 3.2, "entropy": 0.20868069266434758, "epoch": 0.03205128205128205, "frac_reward_zero_std": 0.003125, "grad_norm": 0.5534481406211853, "learning_rate": 1.4893617021276596e-06, "loss": 0.0053, "num_tokens": 23012406.0, "reward": 0.76904296875, "reward_std": 0.8026323795318604, "rewards/accuracy_reward/mean": 0.234375, "rewards/accuracy_reward/std": 0.4228599488735199, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.53466796875, "rewards/format_reward/std": 0.49821196794509887, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 15, "step_time": 278.2607453274337 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02470703125, "completions/max_length": 4096.0, "completions/max_terminated_length": 3986.0, "completions/mean_length": 609.7712890625, "completions/mean_terminated_length": 521.5465026855469, "completions/min_length": 4.8, "completions/min_terminated_length": 4.8, "entropy": 0.21470238340552897, "epoch": 0.042735042735042736, "frac_reward_zero_std": 0.003125, "grad_norm": 2.396941661834717, "learning_rate": 2.021276595744681e-06, "loss": 0.0026, "num_tokens": 30970544.0, "reward": 0.72412109375, "reward_std": 0.783198893070221, "rewards/accuracy_reward/mean": 0.207421875, "rewards/accuracy_reward/std": 0.4049464583396912, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.51669921875, "rewards/format_reward/std": 0.4993353486061096, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 20, "step_time": 281.8500593906589 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0216796875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4000.0, "completions/mean_length": 572.19052734375, "completions/mean_terminated_length": 494.154833984375, "completions/min_length": 4.4, "completions/min_terminated_length": 4.4, "entropy": 0.1887929852702655, "epoch": 0.053418803418803416, "frac_reward_zero_std": 0.003125, "grad_norm": 85.9137954711914, "learning_rate": 2.553191489361702e-06, "loss": 0.0031, "num_tokens": 38500399.0, "reward": 0.7791015625, "reward_std": 0.7995866417884827, "rewards/accuracy_reward/mean": 0.2345703125, "rewards/accuracy_reward/std": 0.4226407825946808, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.54453125, "rewards/format_reward/std": 0.4977679193019867, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 25, "step_time": 277.29182202801167 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02431640625, "completions/max_length": 4096.0, "completions/max_terminated_length": 3974.4, "completions/mean_length": 602.47724609375, "completions/mean_terminated_length": 515.40244140625, "completions/min_length": 5.6, "completions/min_terminated_length": 5.6, "entropy": 0.09899503729830031, "epoch": 0.0641025641025641, "frac_reward_zero_std": 0.0, "grad_norm": 215.24990844726562, "learning_rate": 3.0851063829787237e-06, "loss": 0.007, "num_tokens": 46418982.0, "reward": 0.7408203125, "reward_std": 0.8093600869178772, "rewards/accuracy_reward/mean": 0.2322265625, "rewards/accuracy_reward/std": 0.42136730551719664, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.50859375, "rewards/format_reward/std": 0.49978559613227846, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 30, "step_time": 281.7922251022625 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02392578125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4045.6, "completions/mean_length": 591.63798828125, "completions/mean_terminated_length": 505.85712280273435, "completions/min_length": 3.2, "completions/min_terminated_length": 3.2, "entropy": 0.12875012291042368, "epoch": 0.07478632478632478, "frac_reward_zero_std": 0.003125, "grad_norm": 196.4109344482422, "learning_rate": 3.6170212765957453e-06, "loss": 0.0047, "num_tokens": 54159915.0, "reward": 0.76728515625, "reward_std": 0.7970635533332825, "rewards/accuracy_reward/mean": 0.22890625, "rewards/accuracy_reward/std": 0.4194847583770752, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.53837890625, "rewards/format_reward/std": 0.4982650220394135, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 35, "step_time": 285.2140341614228 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.021875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4018.2, "completions/mean_length": 576.571484375, "completions/mean_terminated_length": 497.8513610839844, "completions/min_length": 3.2, "completions/min_terminated_length": 3.2, "entropy": 0.13915122590842657, "epoch": 0.08547008547008547, "frac_reward_zero_std": 0.0, "grad_norm": 97.68031311035156, "learning_rate": 4.148936170212766e-06, "loss": 0.0052, "num_tokens": 61719719.0, "reward": 0.8001953125, "reward_std": 0.8091744184494019, "rewards/accuracy_reward/mean": 0.24765625, "rewards/accuracy_reward/std": 0.43121981620788574, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.5525390625, "rewards/format_reward/std": 0.49729637503623964, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 40, "step_time": 282.3295673634246 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0234375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4012.6, "completions/mean_length": 595.1443359375, "completions/mean_terminated_length": 511.0856079101562, "completions/min_length": 4.4, "completions/min_terminated_length": 4.4, "entropy": 0.12724743547150866, "epoch": 0.09615384615384616, "frac_reward_zero_std": 0.003125, "grad_norm": 108.57718658447266, "learning_rate": 4.680851063829788e-06, "loss": 0.0046, "num_tokens": 69537965.0, "reward": 0.74765625, "reward_std": 0.8107985377311706, "rewards/accuracy_reward/mean": 0.23466796875, "rewards/accuracy_reward/std": 0.42350680828094484, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.51298828125, "rewards/format_reward/std": 0.49971628189086914, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 45, "step_time": 281.7140337219811 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0240234375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4052.6, "completions/mean_length": 610.7671875, "completions/mean_terminated_length": 525.040869140625, "completions/min_length": 3.4, "completions/min_terminated_length": 3.4, "entropy": 0.12357340005401056, "epoch": 0.10683760683760683, "frac_reward_zero_std": 0.009375, "grad_norm": 74.49510955810547, "learning_rate": 4.9465240641711236e-06, "loss": 0.0053, "num_tokens": 77513885.0, "reward": 0.711328125, "reward_std": 0.7983763813972473, "rewards/accuracy_reward/mean": 0.21669921875, "rewards/accuracy_reward/std": 0.41054518818855285, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.49462890625, "rewards/format_reward/std": 0.4997918665409088, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 50, "step_time": 278.2277604050061 }, { "epoch": 0.10683760683760683, "eval_clip_ratio/high_max": 0.0, "eval_clip_ratio/high_mean": 0.0, "eval_clip_ratio/low_mean": 0.0, "eval_clip_ratio/low_min": 0.0, "eval_clip_ratio/region_mean": 0.0, "eval_completions/clipped_ratio": 0.029296875, "eval_completions/max_length": 3256.375, "eval_completions/max_terminated_length": 1776.28125, "eval_completions/mean_length": 618.1806640625, "eval_completions/mean_terminated_length": 512.6767168045044, "eval_completions/min_length": 96.6875, "eval_completions/min_terminated_length": 96.6875, "eval_entropy": 0.13950487226247787, "eval_frac_reward_zero_std": 1.0, "eval_loss": 0.0, "eval_num_tokens": 77513885.0, "eval_reward": 0.76171875, "eval_reward_std": 0.8027244340628386, "eval_rewards/accuracy_reward/mean": 0.232421875, "eval_rewards/accuracy_reward/std": 0.42211480252444744, "eval_rewards/brier_reward/mean": 0.0, "eval_rewards/brier_reward/std": 0.0, "eval_rewards/confidence_one_or_zero/mean": 0.0, "eval_rewards/confidence_one_or_zero/std": 0.0, "eval_rewards/format_reward/mean": 0.529296875, "eval_rewards/format_reward/std": 0.4974179146811366, "eval_rewards/mean_confidence_reward/mean": 0.0, "eval_rewards/mean_confidence_reward/std": 0.0, "eval_runtime": 1045.0286, "eval_samples_per_second": 0.957, "eval_steps_per_second": 0.031, "step": 50 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02001953125, "completions/max_length": 4096.0, "completions/max_terminated_length": 3962.2, "completions/mean_length": 585.74736328125, "completions/mean_terminated_length": 514.0597595214844, "completions/min_length": 3.2, "completions/min_terminated_length": 3.2, "entropy": 0.1371245729562361, "epoch": 0.11752136752136752, "frac_reward_zero_std": 0.003125, "grad_norm": 60.45952224731445, "learning_rate": 4.812834224598931e-06, "loss": 0.0017, "num_tokens": 85182946.0, "reward": 0.7728515625, "reward_std": 0.796334159374237, "rewards/accuracy_reward/mean": 0.230078125, "rewards/accuracy_reward/std": 0.42046673893928527, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.5427734375, "rewards/format_reward/std": 0.4974696457386017, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 55, "step_time": 279.256950871172 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02353515625, "completions/max_length": 4096.0, "completions/max_terminated_length": 3948.8, "completions/mean_length": 598.4044921875, "completions/mean_terminated_length": 514.1934936523437, "completions/min_length": 4.6, "completions/min_terminated_length": 4.6, "entropy": 0.13020556099363603, "epoch": 0.1282051282051282, "frac_reward_zero_std": 0.003125, "grad_norm": 44.859046936035156, "learning_rate": 4.6791443850267385e-06, "loss": 0.0049, "num_tokens": 93001872.0, "reward": 0.7390625, "reward_std": 0.7925039291381836, "rewards/accuracy_reward/mean": 0.21875, "rewards/accuracy_reward/std": 0.41299651861190795, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.5203125, "rewards/format_reward/std": 0.4984662115573883, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 60, "step_time": 289.76228219861224 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02568359375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4019.4, "completions/mean_length": 624.25341796875, "completions/mean_terminated_length": 532.8375122070313, "completions/min_length": 4.0, "completions/min_terminated_length": 4.0, "entropy": 0.16305259802902583, "epoch": 0.1388888888888889, "frac_reward_zero_std": 0.0, "grad_norm": 43.33089065551758, "learning_rate": 4.5454545454545455e-06, "loss": 0.0039, "num_tokens": 101065843.0, "reward": 0.7591796875, "reward_std": 0.7913376212120056, "rewards/accuracy_reward/mean": 0.2240234375, "rewards/accuracy_reward/std": 0.4148717701435089, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.53515625, "rewards/format_reward/std": 0.49778724312782285, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 65, "step_time": 320.13026094478556 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.023828125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4005.4, "completions/mean_length": 608.73076171875, "completions/mean_terminated_length": 523.7735900878906, "completions/min_length": 5.2, "completions/min_terminated_length": 5.2, "entropy": 0.17717687981785274, "epoch": 0.14957264957264957, "frac_reward_zero_std": 0.0, "grad_norm": 103.29003143310547, "learning_rate": 4.411764705882353e-06, "loss": 0.0041, "num_tokens": 109003310.0, "reward": 0.72451171875, "reward_std": 0.7816438794136047, "rewards/accuracy_reward/mean": 0.2080078125, "rewards/accuracy_reward/std": 0.4034444451332092, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.51650390625, "rewards/format_reward/std": 0.4989802181720734, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 70, "step_time": 283.5916655063658 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0234375, "completions/max_length": 4096.0, "completions/max_terminated_length": 3987.0, "completions/mean_length": 606.1939453125, "completions/mean_terminated_length": 522.5233459472656, "completions/min_length": 3.4, "completions/min_terminated_length": 3.4, "entropy": 0.123387516773073, "epoch": 0.16025641025641027, "frac_reward_zero_std": 0.0, "grad_norm": 38.79195022583008, "learning_rate": 4.2780748663101604e-06, "loss": 0.0046, "num_tokens": 116878608.0, "reward": 0.7271484375, "reward_std": 0.8021309733390808, "rewards/accuracy_reward/mean": 0.22294921875, "rewards/accuracy_reward/std": 0.41599671840667723, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.50419921875, "rewards/format_reward/std": 0.49959311485290525, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 75, "step_time": 284.3083988328232 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0248046875, "completions/max_length": 4096.0, "completions/max_terminated_length": 3953.8, "completions/mean_length": 590.29609375, "completions/mean_terminated_length": 501.14326171875, "completions/min_length": 5.6, "completions/min_terminated_length": 5.6, "entropy": 0.11684439122036565, "epoch": 0.17094017094017094, "frac_reward_zero_std": 0.003125, "grad_norm": 49.576934814453125, "learning_rate": 4.144385026737968e-06, "loss": 0.0029, "num_tokens": 124602696.0, "reward": 0.7560546875, "reward_std": 0.7941526412963867, "rewards/accuracy_reward/mean": 0.22333984375, "rewards/accuracy_reward/std": 0.41590661406517027, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.53271484375, "rewards/format_reward/std": 0.4989008605480194, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 80, "step_time": 285.54715362380665 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0251953125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4000.2, "completions/mean_length": 605.1615234375, "completions/mean_terminated_length": 514.9618225097656, "completions/min_length": 2.4, "completions/min_terminated_length": 2.4, "entropy": 0.11611211430863477, "epoch": 0.18162393162393162, "frac_reward_zero_std": 0.0, "grad_norm": 37.76605987548828, "learning_rate": 4.010695187165775e-06, "loss": 0.0054, "num_tokens": 132457662.0, "reward": 0.75517578125, "reward_std": 0.8076836228370666, "rewards/accuracy_reward/mean": 0.2341796875, "rewards/accuracy_reward/std": 0.4225840926170349, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.52099609375, "rewards/format_reward/std": 0.49966921210289, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 85, "step_time": 279.26752776298673 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02509765625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4012.6, "completions/mean_length": 605.06943359375, "completions/mean_terminated_length": 515.3127136230469, "completions/min_length": 4.8, "completions/min_terminated_length": 4.8, "entropy": 0.1194009205326438, "epoch": 0.19230769230769232, "frac_reward_zero_std": 0.003125, "grad_norm": 36.28676223754883, "learning_rate": 3.877005347593583e-06, "loss": 0.0046, "num_tokens": 140304261.0, "reward": 0.75048828125, "reward_std": 0.7870240211486816, "rewards/accuracy_reward/mean": 0.216796875, "rewards/accuracy_reward/std": 0.4116081833839417, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.53369140625, "rewards/format_reward/std": 0.49821258187294004, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 90, "step_time": 282.06346071061853 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02314453125, "completions/max_length": 4096.0, "completions/max_terminated_length": 3985.0, "completions/mean_length": 594.9970703125, "completions/mean_terminated_length": 512.0561096191407, "completions/min_length": 2.4, "completions/min_terminated_length": 2.4, "entropy": 0.12444208567030728, "epoch": 0.202991452991453, "frac_reward_zero_std": 0.003125, "grad_norm": 35.040470123291016, "learning_rate": 3.7433155080213907e-06, "loss": 0.005, "num_tokens": 148078855.0, "reward": 0.7634765625, "reward_std": 0.8031808733940125, "rewards/accuracy_reward/mean": 0.2326171875, "rewards/accuracy_reward/std": 0.42190982699394225, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.530859375, "rewards/format_reward/std": 0.49894219636917114, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 95, "step_time": 276.9192532343703 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0228515625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4000.4, "completions/mean_length": 577.3484375, "completions/mean_terminated_length": 495.12913208007814, "completions/min_length": 3.2, "completions/min_terminated_length": 3.2, "entropy": 0.13522679468151183, "epoch": 0.21367521367521367, "frac_reward_zero_std": 0.003125, "grad_norm": 38.252925872802734, "learning_rate": 3.609625668449198e-06, "loss": 0.0093, "num_tokens": 155659447.0, "reward": 0.80048828125, "reward_std": 0.8131750822067261, "rewards/accuracy_reward/mean": 0.25126953125, "rewards/accuracy_reward/std": 0.43329434990882876, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.54921875, "rewards/format_reward/std": 0.4973772406578064, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 100, "step_time": 279.8197010210104 }, { "epoch": 0.21367521367521367, "eval_clip_ratio/high_max": 0.0, "eval_clip_ratio/high_mean": 0.0, "eval_clip_ratio/low_mean": 0.0, "eval_clip_ratio/low_min": 0.0, "eval_clip_ratio/region_mean": 0.0, "eval_completions/clipped_ratio": 0.013671875, "eval_completions/max_length": 2739.78125, "eval_completions/max_terminated_length": 1980.53125, "eval_completions/mean_length": 571.171875, "eval_completions/mean_terminated_length": 522.2637557983398, "eval_completions/min_length": 100.71875, "eval_completions/min_terminated_length": 100.71875, "eval_entropy": 0.12239723512902856, "eval_frac_reward_zero_std": 1.0, "eval_loss": 0.0, "eval_num_tokens": 155659447.0, "eval_reward": 0.771484375, "eval_reward_std": 0.7972938418388367, "eval_rewards/accuracy_reward/mean": 0.2294921875, "eval_rewards/accuracy_reward/std": 0.41721432469785213, "eval_rewards/brier_reward/mean": 0.0, "eval_rewards/brier_reward/std": 0.0, "eval_rewards/confidence_one_or_zero/mean": 0.0, "eval_rewards/confidence_one_or_zero/std": 0.0, "eval_rewards/format_reward/mean": 0.5419921875, "eval_rewards/format_reward/std": 0.4998957645148039, "eval_rewards/mean_confidence_reward/mean": 0.0, "eval_rewards/mean_confidence_reward/std": 0.0, "eval_runtime": 948.9157, "eval_samples_per_second": 1.054, "eval_steps_per_second": 0.034, "step": 100 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02216796875, "completions/max_length": 4096.0, "completions/max_terminated_length": 3854.0, "completions/mean_length": 582.88525390625, "completions/mean_terminated_length": 503.2716125488281, "completions/min_length": 4.8, "completions/min_terminated_length": 4.8, "entropy": 0.13384314943104983, "epoch": 0.22435897435897437, "frac_reward_zero_std": 0.009375, "grad_norm": 52.33734893798828, "learning_rate": 3.4759358288770056e-06, "loss": 0.0067, "num_tokens": 163298240.0, "reward": 0.755078125, "reward_std": 0.7937467455863952, "rewards/accuracy_reward/mean": 0.22412109375, "rewards/accuracy_reward/std": 0.41622872948646544, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.53095703125, "rewards/format_reward/std": 0.49783012866973875, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 105, "step_time": 278.9125811098318 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0201171875, "completions/max_length": 4096.0, "completions/max_terminated_length": 3960.0, "completions/mean_length": 579.79892578125, "completions/mean_terminated_length": 507.60706787109376, "completions/min_length": 5.0, "completions/min_terminated_length": 5.0, "entropy": 0.12177688683441375, "epoch": 0.23504273504273504, "frac_reward_zero_std": 0.0, "grad_norm": 45.7343635559082, "learning_rate": 3.342245989304813e-06, "loss": 0.0062, "num_tokens": 170921269.0, "reward": 0.7494140625, "reward_std": 0.7981451392173767, "rewards/accuracy_reward/mean": 0.2248046875, "rewards/accuracy_reward/std": 0.4169042706489563, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.524609375, "rewards/format_reward/std": 0.4992986679077148, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 110, "step_time": 290.9905573850498 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.023828125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4020.4, "completions/mean_length": 594.2623046875, "completions/mean_terminated_length": 508.7661437988281, "completions/min_length": 3.2, "completions/min_terminated_length": 3.2, "entropy": 0.11783788996108342, "epoch": 0.24572649572649571, "frac_reward_zero_std": 0.009375, "grad_norm": 33.649715423583984, "learning_rate": 3.2085561497326205e-06, "loss": 0.0041, "num_tokens": 178683731.0, "reward": 0.7380859375, "reward_std": 0.8020292282104492, "rewards/accuracy_reward/mean": 0.2255859375, "rewards/accuracy_reward/std": 0.4174380123615265, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.5125, "rewards/format_reward/std": 0.49958932399749756, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 115, "step_time": 281.84625020143284 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02255859375, "completions/max_length": 4096.0, "completions/max_terminated_length": 3999.4, "completions/mean_length": 602.1177734375, "completions/mean_terminated_length": 521.7236328125, "completions/min_length": 4.4, "completions/min_terminated_length": 4.4, "entropy": 0.1267630995673244, "epoch": 0.2564102564102564, "frac_reward_zero_std": 0.009375, "grad_norm": 38.76011276245117, "learning_rate": 3.074866310160428e-06, "loss": 0.0022, "num_tokens": 186535689.0, "reward": 0.7390625, "reward_std": 0.7887951493263244, "rewards/accuracy_reward/mean": 0.2150390625, "rewards/accuracy_reward/std": 0.4107056140899658, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.5240234375, "rewards/format_reward/std": 0.49912756085395815, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 120, "step_time": 282.3706236787839 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02138671875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4042.0, "completions/mean_length": 595.8607421875, "completions/mean_terminated_length": 519.3848327636719, "completions/min_length": 4.8, "completions/min_terminated_length": 4.8, "entropy": 0.12497035869164393, "epoch": 0.2670940170940171, "frac_reward_zero_std": 0.00625, "grad_norm": 36.49787139892578, "learning_rate": 2.9411764705882355e-06, "loss": 0.0069, "num_tokens": 194323159.0, "reward": 0.74248046875, "reward_std": 0.8022560358047486, "rewards/accuracy_reward/mean": 0.2263671875, "rewards/accuracy_reward/std": 0.41835899353027345, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.51611328125, "rewards/format_reward/std": 0.4994750082492828, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 125, "step_time": 298.1305276590254 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.022265625, "completions/max_length": 4096.0, "completions/max_terminated_length": 3964.0, "completions/mean_length": 580.27783203125, "completions/mean_terminated_length": 500.289111328125, "completions/min_length": 2.0, "completions/min_terminated_length": 2.0, "entropy": 0.13873413887922653, "epoch": 0.2777777777777778, "frac_reward_zero_std": 0.003125, "grad_norm": 38.390708923339844, "learning_rate": 2.807486631016043e-06, "loss": 0.0024, "num_tokens": 201946260.0, "reward": 0.77841796875, "reward_std": 0.7998116850852967, "rewards/accuracy_reward/mean": 0.23505859375, "rewards/accuracy_reward/std": 0.423546290397644, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.543359375, "rewards/format_reward/std": 0.49665730595588686, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 130, "step_time": 280.3890097280382 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.01796875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4067.8, "completions/mean_length": 556.99736328125, "completions/mean_terminated_length": 492.2548522949219, "completions/min_length": 4.0, "completions/min_terminated_length": 4.0, "entropy": 0.132820147217717, "epoch": 0.28846153846153844, "frac_reward_zero_std": 0.0, "grad_norm": 35.38172149658203, "learning_rate": 2.673796791443851e-06, "loss": 0.0041, "num_tokens": 209321561.0, "reward": 0.75947265625, "reward_std": 0.7935594916343689, "rewards/accuracy_reward/mean": 0.22392578125, "rewards/accuracy_reward/std": 0.4162346839904785, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.535546875, "rewards/format_reward/std": 0.49860507249832153, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 135, "step_time": 280.2609712522244 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0201171875, "completions/max_length": 4096.0, "completions/max_terminated_length": 3871.8, "completions/mean_length": 577.14951171875, "completions/mean_terminated_length": 505.0685791015625, "completions/min_length": 3.8, "completions/min_terminated_length": 3.8, "entropy": 0.1271175751258852, "epoch": 0.29914529914529914, "frac_reward_zero_std": 0.0, "grad_norm": 35.13084030151367, "learning_rate": 2.5401069518716583e-06, "loss": 0.0032, "num_tokens": 216951732.0, "reward": 0.7724609375, "reward_std": 0.8079434990882873, "rewards/accuracy_reward/mean": 0.23974609375, "rewards/accuracy_reward/std": 0.42588251233100893, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.53271484375, "rewards/format_reward/std": 0.4982447803020477, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 140, "step_time": 279.7070125473576 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02490234375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4055.6, "completions/mean_length": 609.0548828125, "completions/mean_terminated_length": 520.1921447753906, "completions/min_length": 3.4, "completions/min_terminated_length": 3.4, "entropy": 0.12080801308620721, "epoch": 0.30982905982905984, "frac_reward_zero_std": 0.0, "grad_norm": 41.06476974487305, "learning_rate": 2.4064171122994653e-06, "loss": 0.0019, "num_tokens": 224897990.0, "reward": 0.74248046875, "reward_std": 0.79608633518219, "rewards/accuracy_reward/mean": 0.22265625, "rewards/accuracy_reward/std": 0.41532952189445493, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.51982421875, "rewards/format_reward/std": 0.49849451184272764, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 145, "step_time": 282.20083903438064 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.019921875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4028.8, "completions/mean_length": 573.9537109375, "completions/mean_terminated_length": 502.3598876953125, "completions/min_length": 3.6, "completions/min_terminated_length": 3.6, "entropy": 0.13020354545442386, "epoch": 0.32051282051282054, "frac_reward_zero_std": 0.0, "grad_norm": 33.12124252319336, "learning_rate": 2.2727272727272728e-06, "loss": 0.0081, "num_tokens": 232458316.0, "reward": 0.77626953125, "reward_std": 0.7912535548210144, "rewards/accuracy_reward/mean": 0.2263671875, "rewards/accuracy_reward/std": 0.41820659041404723, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.54990234375, "rewards/format_reward/std": 0.49754672646522524, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 150, "step_time": 277.7435175840219 }, { "epoch": 0.32051282051282054, "eval_clip_ratio/high_max": 0.0, "eval_clip_ratio/high_mean": 0.0, "eval_clip_ratio/low_mean": 0.0, "eval_clip_ratio/low_min": 0.0, "eval_clip_ratio/region_mean": 0.0, "eval_completions/clipped_ratio": 0.021484375, "eval_completions/max_length": 3091.53125, "eval_completions/max_terminated_length": 2176.6875, "eval_completions/mean_length": 601.455078125, "eval_completions/mean_terminated_length": 524.5443887710571, "eval_completions/min_length": 88.65625, "eval_completions/min_terminated_length": 88.65625, "eval_entropy": 0.1157567179761827, "eval_frac_reward_zero_std": 1.0, "eval_loss": 0.0, "eval_num_tokens": 232458316.0, "eval_reward": 0.7412109375, "eval_reward_std": 0.7956308983266354, "eval_rewards/accuracy_reward/mean": 0.2177734375, "eval_rewards/accuracy_reward/std": 0.4122256711125374, "eval_rewards/brier_reward/mean": 0.0, "eval_rewards/brier_reward/std": 0.0, "eval_rewards/confidence_one_or_zero/mean": 0.0, "eval_rewards/confidence_one_or_zero/std": 0.0, "eval_rewards/format_reward/mean": 0.5234375, "eval_rewards/format_reward/std": 0.5025244373828173, "eval_rewards/mean_confidence_reward/mean": 0.0, "eval_rewards/mean_confidence_reward/std": 0.0, "eval_runtime": 1036.1942, "eval_samples_per_second": 0.965, "eval_steps_per_second": 0.031, "step": 150 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.01884765625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4000.6, "completions/mean_length": 551.7111328125, "completions/mean_terminated_length": 483.6562561035156, "completions/min_length": 2.8, "completions/min_terminated_length": 2.8, "entropy": 0.13453883749316448, "epoch": 0.3311965811965812, "frac_reward_zero_std": 0.003125, "grad_norm": 34.161277770996094, "learning_rate": 2.1390374331550802e-06, "loss": 0.0056, "num_tokens": 239794430.0, "reward": 0.76982421875, "reward_std": 0.7878788471221924, "rewards/accuracy_reward/mean": 0.22236328125, "rewards/accuracy_reward/std": 0.41579753160476685, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 9.765625e-05, "rewards/confidence_one_or_zero/std": 0.004419417306780815, "rewards/format_reward/mean": 0.54736328125, "rewards/format_reward/std": 0.4969615817070007, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 155, "step_time": 279.48853852035825 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0197265625, "completions/max_length": 4096.0, "completions/max_terminated_length": 3989.6, "completions/mean_length": 556.93115234375, "completions/mean_terminated_length": 485.7301940917969, "completions/min_length": 5.0, "completions/min_terminated_length": 5.0, "entropy": 0.13627081344893668, "epoch": 0.3418803418803419, "frac_reward_zero_std": 0.003125, "grad_norm": 34.00373840332031, "learning_rate": 2.0053475935828877e-06, "loss": 0.0052, "num_tokens": 247165885.0, "reward": 0.800390625, "reward_std": 0.8137112021446228, "rewards/accuracy_reward/mean": 0.2521484375, "rewards/accuracy_reward/std": 0.4331360340118408, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.5482421875, "rewards/format_reward/std": 0.49726054072380066, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 160, "step_time": 278.08321814095837 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.01953125, "completions/max_length": 4096.0, "completions/max_terminated_length": 4013.6, "completions/mean_length": 566.95703125, "completions/mean_terminated_length": 496.7828125, "completions/min_length": 4.2, "completions/min_terminated_length": 4.2, "entropy": 0.15073619410395622, "epoch": 0.3525641025641026, "frac_reward_zero_std": 0.0, "grad_norm": 339.44744873046875, "learning_rate": 1.8716577540106954e-06, "loss": 0.0031, "num_tokens": 254657573.0, "reward": 0.7720703125, "reward_std": 0.8031235098838806, "rewards/accuracy_reward/mean": 0.23642578125, "rewards/accuracy_reward/std": 0.4232724666595459, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.53564453125, "rewards/format_reward/std": 0.4978670537471771, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 165, "step_time": 275.5425803271995 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02470703125, "completions/max_length": 4096.0, "completions/max_terminated_length": 3995.0, "completions/mean_length": 612.71416015625, "completions/mean_terminated_length": 524.5241455078125, "completions/min_length": 5.4, "completions/min_terminated_length": 5.4, "entropy": 0.14152835492277518, "epoch": 0.36324786324786323, "frac_reward_zero_std": 0.0, "grad_norm": 130.8936309814453, "learning_rate": 1.7379679144385028e-06, "loss": 0.0031, "num_tokens": 262621878.0, "reward": 0.75537109375, "reward_std": 0.8010072469711303, "rewards/accuracy_reward/mean": 0.23134765625, "rewards/accuracy_reward/std": 0.41914269924163816, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.5240234375, "rewards/format_reward/std": 0.49830247163772584, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 170, "step_time": 282.48500978334926 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.020703125, "completions/max_length": 4096.0, "completions/max_terminated_length": 3847.6, "completions/mean_length": 570.1248046875, "completions/mean_terminated_length": 495.6341918945312, "completions/min_length": 5.8, "completions/min_terminated_length": 5.8, "entropy": 0.1313963848806452, "epoch": 0.37393162393162394, "frac_reward_zero_std": 0.009375, "grad_norm": 102.34068298339844, "learning_rate": 1.6042780748663103e-06, "loss": 0.0011, "num_tokens": 270156980.0, "reward": 0.79228515625, "reward_std": 0.8168907523155212, "rewards/accuracy_reward/mean": 0.2515625, "rewards/accuracy_reward/std": 0.4333960711956024, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.54072265625, "rewards/format_reward/std": 0.498195743560791, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 175, "step_time": 279.8141848627653 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.01982421875, "completions/max_length": 4096.0, "completions/max_terminated_length": 3991.2, "completions/mean_length": 563.10390625, "completions/mean_terminated_length": 491.7077697753906, "completions/min_length": 4.0, "completions/min_terminated_length": 4.0, "entropy": 0.1302239095501136, "epoch": 0.38461538461538464, "frac_reward_zero_std": 0.0, "grad_norm": 82.37760162353516, "learning_rate": 1.4705882352941177e-06, "loss": 0.0046, "num_tokens": 277607196.0, "reward": 0.76435546875, "reward_std": 0.7951547741889954, "rewards/accuracy_reward/mean": 0.2291015625, "rewards/accuracy_reward/std": 0.41797728538513185, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.53525390625, "rewards/format_reward/std": 0.4972443461418152, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 180, "step_time": 280.01061920716313 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02490234375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4004.0, "completions/mean_length": 607.22412109375, "completions/mean_terminated_length": 518.144775390625, "completions/min_length": 2.6, "completions/min_terminated_length": 2.6, "entropy": 0.12305268299533054, "epoch": 0.3952991452991453, "frac_reward_zero_std": 0.003125, "grad_norm": 63.95575714111328, "learning_rate": 1.3368983957219254e-06, "loss": 0.0031, "num_tokens": 285559923.0, "reward": 0.72919921875, "reward_std": 0.7951399564743042, "rewards/accuracy_reward/mean": 0.21904296875, "rewards/accuracy_reward/std": 0.4109188914299011, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.51015625, "rewards/format_reward/std": 0.4992732465267181, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 185, "step_time": 298.1623516379565 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0220703125, "completions/max_length": 4096.0, "completions/max_terminated_length": 3711.4, "completions/mean_length": 597.7697265625, "completions/mean_terminated_length": 519.1280700683594, "completions/min_length": 5.6, "completions/min_terminated_length": 5.6, "entropy": 0.1266944181465078, "epoch": 0.405982905982906, "frac_reward_zero_std": 0.003125, "grad_norm": 53.50038528442383, "learning_rate": 1.2032085561497326e-06, "loss": 0.0038, "num_tokens": 293355613.0, "reward": 0.73486328125, "reward_std": 0.7907920479774475, "rewards/accuracy_reward/mean": 0.21669921875, "rewards/accuracy_reward/std": 0.4108963131904602, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.5181640625, "rewards/format_reward/std": 0.49884843826293945, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 190, "step_time": 280.70746657224373 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02578125, "completions/max_length": 4096.0, "completions/max_terminated_length": 3949.2, "completions/mean_length": 604.76962890625, "completions/mean_terminated_length": 512.4957946777344, "completions/min_length": 4.0, "completions/min_terminated_length": 4.0, "entropy": 0.13020834482740612, "epoch": 0.4166666666666667, "frac_reward_zero_std": 0.0, "grad_norm": 71.60282135009766, "learning_rate": 1.0695187165775401e-06, "loss": 0.0038, "num_tokens": 301250694.0, "reward": 0.73896484375, "reward_std": 0.800883674621582, "rewards/accuracy_reward/mean": 0.22529296875, "rewards/accuracy_reward/std": 0.41599961519241335, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.513671875, "rewards/format_reward/std": 0.4998628258705139, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 195, "step_time": 280.38385073145037 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.024609375, "completions/max_length": 4096.0, "completions/max_terminated_length": 3951.4, "completions/mean_length": 599.652734375, "completions/mean_terminated_length": 511.47874755859374, "completions/min_length": 4.8, "completions/min_terminated_length": 4.8, "entropy": 0.13503547279688063, "epoch": 0.42735042735042733, "frac_reward_zero_std": 0.003125, "grad_norm": 45.0015869140625, "learning_rate": 9.358288770053477e-07, "loss": 0.0042, "num_tokens": 309072386.0, "reward": 0.75546875, "reward_std": 0.7948607087135315, "rewards/accuracy_reward/mean": 0.22373046875, "rewards/accuracy_reward/std": 0.41643038392066956, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.53173828125, "rewards/format_reward/std": 0.4988611042499542, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 200, "step_time": 280.60323548226154 }, { "epoch": 0.42735042735042733, "eval_clip_ratio/high_max": 0.0, "eval_clip_ratio/high_mean": 0.0, "eval_clip_ratio/low_mean": 0.0, "eval_clip_ratio/low_min": 0.0, "eval_clip_ratio/region_mean": 0.0, "eval_completions/clipped_ratio": 0.0224609375, "eval_completions/max_length": 3271.8125, "eval_completions/max_terminated_length": 2094.75, "eval_completions/mean_length": 602.595703125, "eval_completions/mean_terminated_length": 521.8194351196289, "eval_completions/min_length": 101.28125, "eval_completions/min_terminated_length": 101.28125, "eval_entropy": 0.12412643106654286, "eval_frac_reward_zero_std": 1.0, "eval_loss": 0.0, "eval_num_tokens": 309072386.0, "eval_reward": 0.7666015625, "eval_reward_std": 0.7991905082017183, "eval_rewards/accuracy_reward/mean": 0.2275390625, "eval_rewards/accuracy_reward/std": 0.41939933877438307, "eval_rewards/brier_reward/mean": 0.0, "eval_rewards/brier_reward/std": 0.0, "eval_rewards/confidence_one_or_zero/mean": 0.0, "eval_rewards/confidence_one_or_zero/std": 0.0, "eval_rewards/format_reward/mean": 0.5390625, "eval_rewards/format_reward/std": 0.5004308177158237, "eval_rewards/mean_confidence_reward/mean": 0.0, "eval_rewards/mean_confidence_reward/std": 0.0, "eval_runtime": 1070.3118, "eval_samples_per_second": 0.934, "eval_steps_per_second": 0.03, "step": 200 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.01875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4053.2, "completions/mean_length": 561.92626953125, "completions/mean_terminated_length": 494.40770263671874, "completions/min_length": 4.0, "completions/min_terminated_length": 4.0, "entropy": 0.13653398368041963, "epoch": 0.43803418803418803, "frac_reward_zero_std": 0.0, "grad_norm": 42.5616340637207, "learning_rate": 8.021390374331551e-07, "loss": 0.002, "num_tokens": 316509007.0, "reward": 0.80263671875, "reward_std": 0.8056981444358826, "rewards/accuracy_reward/mean": 0.246484375, "rewards/accuracy_reward/std": 0.4298031270503998, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.55615234375, "rewards/format_reward/std": 0.4965605974197388, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 205, "step_time": 278.53085225895046 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0234375, "completions/max_length": 4096.0, "completions/max_terminated_length": 4030.8, "completions/mean_length": 602.79169921875, "completions/mean_terminated_length": 519.0871826171875, "completions/min_length": 2.8, "completions/min_terminated_length": 2.8, "entropy": 0.13000990800064754, "epoch": 0.44871794871794873, "frac_reward_zero_std": 0.003125, "grad_norm": 158.82310485839844, "learning_rate": 6.684491978609627e-07, "loss": 0.004, "num_tokens": 324384058.0, "reward": 0.75625, "reward_std": 0.7982043027877808, "rewards/accuracy_reward/mean": 0.2279296875, "rewards/accuracy_reward/std": 0.4183617830276489, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.5283203125, "rewards/format_reward/std": 0.49839370250701903, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 210, "step_time": 281.4074709706474 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02265625, "completions/max_length": 4096.0, "completions/max_terminated_length": 4030.8, "completions/mean_length": 592.18095703125, "completions/mean_terminated_length": 510.86790771484374, "completions/min_length": 3.4, "completions/min_terminated_length": 3.4, "entropy": 0.1319914077001158, "epoch": 0.4594017094017094, "frac_reward_zero_std": 0.0, "grad_norm": 38.93572998046875, "learning_rate": 5.347593582887701e-07, "loss": 0.0044, "num_tokens": 332116663.0, "reward": 0.78701171875, "reward_std": 0.8117906808853149, "rewards/accuracy_reward/mean": 0.24609375, "rewards/accuracy_reward/std": 0.4304999113082886, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.54091796875, "rewards/format_reward/std": 0.4979578137397766, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 215, "step_time": 280.5660892641114 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0244140625, "completions/max_length": 4096.0, "completions/max_terminated_length": 3973.6, "completions/mean_length": 596.687109375, "completions/mean_terminated_length": 509.20489501953125, "completions/min_length": 6.6, "completions/min_terminated_length": 6.6, "entropy": 0.13046115188917612, "epoch": 0.4700854700854701, "frac_reward_zero_std": 0.003125, "grad_norm": 79.85475158691406, "learning_rate": 4.0106951871657757e-07, "loss": 0.0059, "num_tokens": 339895283.0, "reward": 0.7552734375, "reward_std": 0.8010589480400085, "rewards/accuracy_reward/mean": 0.22890625, "rewards/accuracy_reward/std": 0.41962975859642027, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.5263671875, "rewards/format_reward/std": 0.49905532598495483, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 220, "step_time": 280.56789546939547 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02216796875, "completions/max_length": 4096.0, "completions/max_terminated_length": 4008.2, "completions/mean_length": 594.92587890625, "completions/mean_terminated_length": 515.61923828125, "completions/min_length": 8.2, "completions/min_terminated_length": 8.2, "entropy": 0.13403512026416137, "epoch": 0.4807692307692308, "frac_reward_zero_std": 0.0, "grad_norm": 43.87257766723633, "learning_rate": 2.6737967914438503e-07, "loss": 0.0044, "num_tokens": 347645596.0, "reward": 0.77333984375, "reward_std": 0.8029205560684204, "rewards/accuracy_reward/mean": 0.23515625, "rewards/accuracy_reward/std": 0.42359659671783445, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.53818359375, "rewards/format_reward/std": 0.4983603596687317, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 225, "step_time": 280.5630801010062 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.02158203125, "completions/max_length": 4096.0, "completions/max_terminated_length": 3956.2, "completions/mean_length": 593.143359375, "completions/mean_terminated_length": 515.9145202636719, "completions/min_length": 5.4, "completions/min_terminated_length": 5.4, "entropy": 0.12892428693012334, "epoch": 0.49145299145299143, "frac_reward_zero_std": 0.00625, "grad_norm": 41.521766662597656, "learning_rate": 1.3368983957219251e-07, "loss": 0.0079, "num_tokens": 355381176.0, "reward": 0.7603515625, "reward_std": 0.8110703587532043, "rewards/accuracy_reward/mean": 0.23857421875, "rewards/accuracy_reward/std": 0.4259735465049744, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.52177734375, "rewards/format_reward/std": 0.49879233837127684, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 230, "step_time": 280.4674930868379 }, { "clip_ratio/high_max": 0.0, "clip_ratio/high_mean": 0.0, "clip_ratio/low_mean": 0.0, "clip_ratio/low_min": 0.0, "clip_ratio/region_mean": 0.0, "completions/clipped_ratio": 0.0205078125, "completions/max_length": 4096.0, "completions/max_terminated_length": 3962.0, "completions/mean_length": 585.1878662109375, "completions/mean_terminated_length": 511.71815490722656, "completions/min_length": 3.75, "completions/min_terminated_length": 3.75, "entropy": 0.1313832072846708, "epoch": 0.5, "frac_reward_zero_std": 0.00390625, "num_tokens": 361528315.0, "reward": 0.803955078125, "reward_std": 0.817492201924324, "rewards/accuracy_reward/mean": 0.25537109375, "rewards/accuracy_reward/std": 0.43593502789735794, "rewards/brier_reward/mean": 0.0, "rewards/brier_reward/std": 0.0, "rewards/confidence_one_or_zero/mean": 0.0, "rewards/confidence_one_or_zero/std": 0.0, "rewards/format_reward/mean": 0.548583984375, "rewards/format_reward/std": 0.49755311757326126, "rewards/mean_confidence_reward/mean": 0.0, "rewards/mean_confidence_reward/std": 0.0, "step": 234, "step_time": 279.6093821763934, "total_flos": 0.0, "train_loss": 0.004653439443144534, "train_runtime": 73926.8948, "train_samples_per_second": 0.203, "train_steps_per_second": 0.003 } ], "logging_steps": 5, "max_steps": 234, "num_input_tokens_seen": 361528315, "num_train_epochs": 1, "save_steps": 60, "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": 4, "trial_name": null, "trial_params": null }