{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "eval_steps": 500, "global_step": 150, "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": 0.0125, "completions/max_length": 454.5, "completions/max_terminated_length": 342.3, "completions/mean_length": 139.7, "completions/mean_terminated_length": 119.76785888671876, "completions/min_length": 11.4, "completions/min_terminated_length": 11.4, "entropy": 0.45251130862161515, "epoch": 0.06666666666666667, "frac_reward_zero_std": 0.55, "grad_norm": 24.625, "kl": 0.003444959384390245, "learning_rate": 9.911436253643443e-07, "loss": -0.056199127435684205, "num_tokens": 22424.0, "reward": 1.4375, "reward_std": 0.44930775463581085, "rewards/accuracy_reward_func/mean": 0.45, "rewards/accuracy_reward_func/std": 0.42546978294849397, "rewards/format_reward_func/mean": 0.9875, "rewards/format_reward_func/std": 0.03535533845424652, "step": 10, "step_time": 17.857542922068387 }, { "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": 0.0, "completions/max_length": 405.7, "completions/max_terminated_length": 405.7, "completions/mean_length": 104.75, "completions/mean_terminated_length": 104.75, "completions/min_length": 11.0, "completions/min_terminated_length": 11.0, "entropy": 0.37909337431192397, "epoch": 0.13333333333333333, "frac_reward_zero_std": 0.6, "grad_norm": 33.25, "kl": 0.0022808335735902572, "learning_rate": 9.609315757942502e-07, "loss": -0.12742023468017577, "num_tokens": 41264.0, "reward": 1.65, "reward_std": 0.369350728392601, "rewards/accuracy_reward_func/mean": 0.675, "rewards/accuracy_reward_func/std": 0.32949312031269073, "rewards/format_reward_func/mean": 0.975, "rewards/format_reward_func/std": 0.046291005611419675, "step": 20, "step_time": 15.968829826638103 }, { "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": 0.0375, "completions/max_length": 770.7, "completions/max_terminated_length": 607.1, "completions/mean_length": 247.05, "completions/mean_terminated_length": 189.3309555053711, "completions/min_length": 11.0, "completions/min_terminated_length": 11.0, "entropy": 0.4646949103684165, "epoch": 0.2, "frac_reward_zero_std": 0.6, "grad_norm": 5.15625, "kl": 0.003183109380734095, "learning_rate": 9.10574604566852e-07, "loss": 0.23813328742980958, "num_tokens": 72036.0, "reward": 1.6125, "reward_std": 0.5259025603532791, "rewards/accuracy_reward_func/mean": 0.675, "rewards/accuracy_reward_func/std": 0.41283679008483887, "rewards/format_reward_func/mean": 0.9375, "rewards/format_reward_func/std": 0.15235702097415924, "step": 30, "step_time": 28.961046700924634 }, { "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": 0.025, "completions/max_length": 651.1, "completions/max_terminated_length": 495.6, "completions/mean_length": 195.975, "completions/mean_terminated_length": 155.1339324951172, "completions/min_length": 12.4, "completions/min_terminated_length": 12.4, "entropy": 0.42483185429591686, "epoch": 0.26666666666666666, "frac_reward_zero_std": 0.5, "grad_norm": 12.0625, "kl": 0.002366386147068633, "learning_rate": 8.422735529643443e-07, "loss": 0.07951252460479737, "num_tokens": 98462.0, "reward": 1.55, "reward_std": 0.5763359367847443, "rewards/accuracy_reward_func/mean": 0.625, "rewards/accuracy_reward_func/std": 0.4401642054319382, "rewards/format_reward_func/mean": 0.925, "rewards/format_reward_func/std": 0.15782093703746797, "step": 40, "step_time": 24.76931145079434 }, { "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": 0.0, "completions/max_length": 212.6, "completions/max_terminated_length": 212.6, "completions/mean_length": 77.7625, "completions/mean_terminated_length": 77.7625, "completions/min_length": 11.4, "completions/min_terminated_length": 11.4, "entropy": 0.5150532013736665, "epoch": 0.3333333333333333, "frac_reward_zero_std": 0.4, "grad_norm": 24.0, "kl": 0.012370214833595128, "learning_rate": 7.590135046865651e-07, "loss": 0.07622140645980835, "num_tokens": 115615.0, "reward": 1.65, "reward_std": 0.3610740154981613, "rewards/accuracy_reward_func/mean": 0.6625, "rewards/accuracy_reward_func/std": 0.32571869492530825, "rewards/format_reward_func/mean": 0.9875, "rewards/format_reward_func/std": 0.03535533845424652, "step": 50, "step_time": 9.345398245193064 }, { "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": 0.0, "completions/max_length": 391.9, "completions/max_terminated_length": 391.9, "completions/mean_length": 140.5125, "completions/mean_terminated_length": 140.5125, "completions/min_length": 10.5, "completions/min_terminated_length": 10.5, "entropy": 0.4693372020497918, "epoch": 0.4, "frac_reward_zero_std": 0.55, "grad_norm": 0.4921875, "kl": 0.003207334756950786, "learning_rate": 6.644333233692916e-07, "loss": -0.014827732741832734, "num_tokens": 137696.0, "reward": 1.6375, "reward_std": 0.37656808495521543, "rewards/accuracy_reward_func/mean": 0.6625, "rewards/accuracy_reward_func/std": 0.3202626138925552, "rewards/format_reward_func/mean": 0.975, "rewards/format_reward_func/std": 0.07071067690849304, "step": 60, "step_time": 15.56080833701417 }, { "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": 0.025, "completions/max_length": 588.2, "completions/max_terminated_length": 532.9, "completions/mean_length": 150.75, "completions/mean_terminated_length": 106.68928833007813, "completions/min_length": 11.1, "completions/min_terminated_length": 11.1, "entropy": 0.4096149187302217, "epoch": 0.4666666666666667, "frac_reward_zero_std": 0.5, "grad_norm": 14.75, "kl": 0.0029691428710791, "learning_rate": 5.626666167821521e-07, "loss": -0.02979586124420166, "num_tokens": 161620.0, "reward": 1.525, "reward_std": 0.48160980343818666, "rewards/accuracy_reward_func/mean": 0.5625, "rewards/accuracy_reward_func/std": 0.4115981698036194, "rewards/format_reward_func/mean": 0.9625, "rewards/format_reward_func/std": 0.10606601536273956, "step": 70, "step_time": 22.58805529791862 }, { "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": 0.0, "completions/max_length": 296.8, "completions/max_terminated_length": 296.8, "completions/mean_length": 94.425, "completions/mean_terminated_length": 94.425, "completions/min_length": 11.5, "completions/min_terminated_length": 11.5, "entropy": 0.3293402023613453, "epoch": 0.5333333333333333, "frac_reward_zero_std": 0.75, "grad_norm": 1.1953125, "kl": 0.0025371614843209046, "learning_rate": 4.5816107833384233e-07, "loss": -0.060630393028259275, "num_tokens": 180298.0, "reward": 1.8, "reward_std": 0.3336638957262039, "rewards/accuracy_reward_func/mean": 0.8125, "rewards/accuracy_reward_func/std": 0.3055525213479996, "rewards/format_reward_func/mean": 0.9875, "rewards/format_reward_func/std": 0.03535533845424652, "step": 80, "step_time": 12.384980474878102 }, { "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": 0.0125, "completions/max_length": 571.0, "completions/max_terminated_length": 396.8, "completions/mean_length": 169.8875, "completions/mean_terminated_length": 145.42857208251954, "completions/min_length": 11.6, "completions/min_terminated_length": 11.6, "entropy": 0.32843760270625355, "epoch": 0.6, "frac_reward_zero_std": 0.75, "grad_norm": 0.265625, "kl": 0.006397408900011214, "learning_rate": 3.554841015277641e-07, "loss": 0.06048367619514465, "num_tokens": 204697.0, "reward": 1.725, "reward_std": 0.2411646068096161, "rewards/accuracy_reward_func/mean": 0.7375, "rewards/accuracy_reward_func/std": 0.20580926835536956, "rewards/format_reward_func/mean": 0.9875, "rewards/format_reward_func/std": 0.03535533845424652, "step": 90, "step_time": 21.92367275925353 }, { "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": 0.0125, "completions/max_length": 561.9, "completions/max_terminated_length": 436.8, "completions/mean_length": 148.8625, "completions/mean_terminated_length": 124.88392944335938, "completions/min_length": 11.8, "completions/min_terminated_length": 11.8, "entropy": 0.385774496383965, "epoch": 0.6666666666666666, "frac_reward_zero_std": 0.7, "grad_norm": 0.05419921875, "kl": 0.006441528908105454, "learning_rate": 2.591231629491423e-07, "loss": -0.09045490622520447, "num_tokens": 227338.0, "reward": 1.6375, "reward_std": 0.4069852590560913, "rewards/accuracy_reward_func/mean": 0.65, "rewards/accuracy_reward_func/std": 0.38802950978279116, "rewards/format_reward_func/mean": 0.9875, "rewards/format_reward_func/std": 0.03535533845424652, "step": 100, "step_time": 21.567665427271276 }, { "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": 0.0, "completions/max_length": 276.2, "completions/max_terminated_length": 276.2, "completions/mean_length": 65.1375, "completions/mean_terminated_length": 65.1375, "completions/min_length": 10.5, "completions/min_terminated_length": 10.5, "entropy": 0.3241632700897753, "epoch": 0.7333333333333333, "frac_reward_zero_std": 0.5, "grad_norm": 0.98046875, "kl": 0.003847465490668345, "learning_rate": 1.7328969800494726e-07, "loss": 0.09001727104187011, "num_tokens": 242669.0, "reward": 1.7375, "reward_std": 0.4489682465791702, "rewards/accuracy_reward_func/mean": 0.75, "rewards/accuracy_reward_func/std": 0.428018119931221, "rewards/format_reward_func/mean": 0.9875, "rewards/format_reward_func/std": 0.03535533845424652, "step": 110, "step_time": 11.50382228847593 }, { "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": 0.0375, "completions/max_length": 851.4, "completions/max_terminated_length": 603.7, "completions/mean_length": 275.25, "completions/mean_terminated_length": 206.3190490722656, "completions/min_length": 11.0, "completions/min_terminated_length": 11.0, "entropy": 0.41810003817081454, "epoch": 0.8, "frac_reward_zero_std": 0.7, "grad_norm": 1.28125, "kl": 0.011894569395826693, "learning_rate": 1.0173504098790186e-07, "loss": 0.12610946893692015, "num_tokens": 275633.0, "reward": 1.6, "reward_std": 0.45666705071926117, "rewards/accuracy_reward_func/mean": 0.6375, "rewards/accuracy_reward_func/std": 0.4069852650165558, "rewards/format_reward_func/mean": 0.9625, "rewards/format_reward_func/std": 0.0816463440656662, "step": 120, "step_time": 31.933318979479374 }, { "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": 0.0125, "completions/max_length": 568.3, "completions/max_terminated_length": 374.9, "completions/mean_length": 130.45, "completions/mean_terminated_length": 105.21250019073486, "completions/min_length": 11.2, "completions/min_terminated_length": 11.2, "entropy": 0.45407755924388765, "epoch": 0.8666666666666667, "frac_reward_zero_std": 0.55, "grad_norm": 51.75, "kl": 0.008640645261036185, "learning_rate": 4.758647376699032e-08, "loss": 0.14615154266357422, "num_tokens": 297813.0, "reward": 1.5625, "reward_std": 0.3699325889348984, "rewards/accuracy_reward_func/mean": 0.5875, "rewards/accuracy_reward_func/std": 0.316480353474617, "rewards/format_reward_func/mean": 0.975, "rewards/format_reward_func/std": 0.07071067690849304, "step": 130, "step_time": 21.87842119038105 }, { "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": 0.0, "completions/max_length": 117.7, "completions/max_terminated_length": 117.7, "completions/mean_length": 32.7375, "completions/mean_terminated_length": 32.7375, "completions/min_length": 10.8, "completions/min_terminated_length": 10.8, "entropy": 0.40678847576491534, "epoch": 0.9333333333333333, "frac_reward_zero_std": 0.7, "grad_norm": 1.03125, "kl": 0.005609498374019495, "learning_rate": 1.3210548563419855e-08, "loss": 0.04860760569572449, "num_tokens": 311608.0, "reward": 1.6, "reward_std": 0.31909499764442445, "rewards/accuracy_reward_func/mean": 0.6125, "rewards/accuracy_reward_func/std": 0.30129902362823485, "rewards/format_reward_func/mean": 0.9875, "rewards/format_reward_func/std": 0.03535533845424652, "step": 140, "step_time": 6.0708080609329045 }, { "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": 0.0, "completions/max_length": 390.7, "completions/max_terminated_length": 390.7, "completions/mean_length": 116.95, "completions/mean_terminated_length": 116.95, "completions/min_length": 10.9, "completions/min_terminated_length": 10.9, "entropy": 0.41855254471302034, "epoch": 1.0, "frac_reward_zero_std": 0.55, "grad_norm": 0.1533203125, "kl": 0.0024387751711611116, "learning_rate": 1.0965826257725019e-10, "loss": -0.009271913766860962, "num_tokens": 332084.0, "reward": 1.4875, "reward_std": 0.4244658529758453, "rewards/accuracy_reward_func/mean": 0.5125, "rewards/accuracy_reward_func/std": 0.36824311017990113, "rewards/format_reward_func/mean": 0.975, "rewards/format_reward_func/std": 0.07071067690849304, "step": 150, "step_time": 15.536289924476296 } ], "logging_steps": 10, "max_steps": 150, "num_input_tokens_seen": 332084, "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": 1, "trial_name": null, "trial_params": null }