{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.9984, "eval_steps": 50, "global_step": 312, "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": 0.10869140625, "completions/max_length": 1536.0, "completions/max_terminated_length": 1530.8, "completions/mean_length": 466.23115234375, "completions/mean_terminated_length": 335.8650329589844, "completions/min_length": 1.0, "completions/min_terminated_length": 1.0, "epoch": 0.016, "grad_norm": 0.003617351409047842, "learning_rate": 3.1249999999999997e-07, "loss": 0.0435, "num_tokens": 19618239.0, "reward": 0.19849815964698792, "reward_std": 0.3928296208381653, "rewards/accuracy_reward": 0.0673828125, "rewards/brier_reward": 0.11236810982227326, "rewards/confidence_one_or_zero": 0.4076171875, "rewards/format_reward": 0.26376953125, "rewards/mean_confidence_reward": 0.5322783350944519, "rewards/smce": -0.4652600884437561, "step": 5 }, { "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": 0.0974609375, "completions/max_length": 1536.0, "completions/max_terminated_length": 1532.8, "completions/mean_length": 438.4470703125, "completions/mean_terminated_length": 320.0117980957031, "completions/min_length": 1.0, "completions/min_terminated_length": 1.0, "epoch": 0.032, "grad_norm": 0.00421218853443861, "learning_rate": 6.249999999999999e-07, "loss": 0.0504, "num_tokens": 39208289.0, "reward": 0.2096415013074875, "reward_std": 0.39408687949180604, "rewards/accuracy_reward": 0.06171875, "rewards/brier_reward": 0.11634623855352402, "rewards/confidence_one_or_zero": 0.4392578125, "rewards/format_reward": 0.2923828125, "rewards/mean_confidence_reward": 0.5730387091636657, "rewards/smce": -0.5116680800914765, "step": 10 }, { "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": 0.05576171875, "completions/max_length": 1536.0, "completions/max_terminated_length": 1515.8, "completions/mean_length": 318.03837890625, "completions/mean_terminated_length": 246.70384826660157, "completions/min_length": 1.0, "completions/min_terminated_length": 1.0, "epoch": 0.048, "grad_norm": 0.004430481232702732, "learning_rate": 9.374999999999999e-07, "loss": 0.0466, "num_tokens": 57513738.0, "reward": 0.37149760127067566, "reward_std": 0.4421126425266266, "rewards/accuracy_reward": 0.10546875, "rewards/brier_reward": 0.1947878420352936, "rewards/confidence_one_or_zero": 0.528125, "rewards/format_reward": 0.5015625, "rewards/mean_confidence_reward": 0.6924709677696228, "rewards/smce": -0.588262677192688, "step": 15 }, { "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": 0.02119140625, "completions/max_length": 1536.0, "completions/max_terminated_length": 1389.4, "completions/mean_length": 209.79072265625, "completions/mean_terminated_length": 181.20720825195312, "completions/min_length": 21.0, "completions/min_terminated_length": 21.0, "epoch": 0.064, "grad_norm": 0.002647513523697853, "learning_rate": 1e-06, "loss": 0.0374, "num_tokens": 74580395.0, "reward": 0.6562158942222596, "reward_std": 0.4088839590549469, "rewards/accuracy_reward": 0.1923828125, "rewards/brier_reward": 0.3430897116661072, "rewards/confidence_one_or_zero": 0.55888671875, "rewards/format_reward": 0.83701171875, "rewards/mean_confidence_reward": 0.7904770016670227, "rewards/smce": -0.6005512475967407, "step": 20 }, { "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": 0.01181640625, "completions/max_length": 1536.0, "completions/max_terminated_length": 1300.0, "completions/mean_length": 161.79404296875, "completions/mean_terminated_length": 145.35975952148436, "completions/min_length": 22.2, "completions/min_terminated_length": 22.2, "epoch": 0.08, "grad_norm": 0.0026642223820090294, "learning_rate": 1e-06, "loss": 0.0175, "num_tokens": 91170318.0, "reward": 0.774070394039154, "reward_std": 0.34370283484458924, "rewards/accuracy_reward": 0.2166015625, "rewards/brier_reward": 0.42684093713760374, "rewards/confidence_one_or_zero": 0.48447265625, "rewards/format_reward": 0.959765625, "rewards/mean_confidence_reward": 0.7623114109039306, "rewards/smce": -0.5506982326507568, "step": 25 }, { "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": 0.00927734375, "completions/max_length": 1536.0, "completions/max_terminated_length": 1212.6, "completions/mean_length": 149.14765625, "completions/mean_terminated_length": 136.1572692871094, "completions/min_length": 25.0, "completions/min_terminated_length": 25.0, "epoch": 0.096, "grad_norm": 0.002164568053558469, "learning_rate": 1e-06, "loss": 0.0148, "num_tokens": 107742198.0, "reward": 0.8744075894355774, "reward_std": 0.29445708990097047, "rewards/accuracy_reward": 0.23955078125, "rewards/brier_reward": 0.5632546305656433, "rewards/confidence_one_or_zero": 0.3158203125, "rewards/format_reward": 0.98310546875, "rewards/mean_confidence_reward": 0.5917321562767028, "rewards/smce": -0.3709741532802582, "step": 30 }, { "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": 0.0052734375, "completions/max_length": 1536.0, "completions/max_terminated_length": 912.6, "completions/mean_length": 133.21044921875, "completions/mean_terminated_length": 125.7759017944336, "completions/min_length": 32.4, "completions/min_terminated_length": 32.4, "epoch": 0.112, "grad_norm": 0.0014111875789240003, "learning_rate": 1e-06, "loss": 0.0105, "num_tokens": 124215745.0, "reward": 0.9913592576980591, "reward_std": 0.20873564779758452, "rewards/accuracy_reward": 0.28134765625, "rewards/brier_reward": 0.7238480806350708, "rewards/confidence_one_or_zero": 0.25, "rewards/format_reward": 0.99072265625, "rewards/mean_confidence_reward": 0.3475369572639465, "rewards/smce": -0.13201046139001846, "step": 35 }, { "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": 0.00126953125, "completions/max_length": 1536.0, "completions/max_terminated_length": 604.0, "completions/mean_length": 123.39462890625, "completions/mean_terminated_length": 121.59987945556641, "completions/min_length": 37.8, "completions/min_terminated_length": 37.8, "epoch": 0.128, "grad_norm": 0.0009458517306484282, "learning_rate": 1e-06, "loss": 0.0029, "num_tokens": 140395978.0, "reward": 1.018831992149353, "reward_std": 0.13349860906600952, "rewards/accuracy_reward": 0.27802734375, "rewards/brier_reward": 0.7702086210250855, "rewards/confidence_one_or_zero": 0.20068359375, "rewards/format_reward": 0.99609375, "rewards/mean_confidence_reward": 0.2456991046667099, "rewards/smce": -0.06666604205965995, "step": 40 }, { "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": 0.00146484375, "completions/max_length": 1536.0, "completions/max_terminated_length": 626.6, "completions/mean_length": 121.8669921875, "completions/mean_terminated_length": 119.79229125976562, "completions/min_length": 47.4, "completions/min_terminated_length": 47.4, "epoch": 0.144, "grad_norm": 0.0008370755822397768, "learning_rate": 1e-06, "loss": 0.0033, "num_tokens": 156594328.0, "reward": 1.0348623275756836, "reward_std": 0.11074777394533157, "rewards/accuracy_reward": 0.3251953125, "rewards/brier_reward": 0.7585193753242493, "rewards/confidence_one_or_zero": 0.1462890625, "rewards/format_reward": 0.997265625, "rewards/mean_confidence_reward": 0.2217879205942154, "rewards/smce": -0.11256366223096848, "step": 45 }, { "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": 0.00068359375, "completions/max_length": 1290.4, "completions/max_terminated_length": 458.2, "completions/mean_length": 122.57265625, "completions/mean_terminated_length": 121.60626678466797, "completions/min_length": 51.6, "completions/min_terminated_length": 51.6, "epoch": 0.16, "grad_norm": 0.0007780106971040368, "learning_rate": 1e-06, "loss": 0.0009, "num_tokens": 172870400.0, "reward": 1.0400798797607422, "reward_std": 0.09919532835483551, "rewards/accuracy_reward": 0.31435546875, "rewards/brier_reward": 0.7744772434234619, "rewards/confidence_one_or_zero": 0.07734375, "rewards/format_reward": 0.9986328125, "rewards/mean_confidence_reward": 0.24838550090789796, "rewards/smce": -0.07306429222226143, "step": 50 }, { "epoch": 0.16, "eval_completions/clipped_ratio": 0.0, "eval_completions/max_length": 278.0, "eval_completions/max_terminated_length": 278.0, "eval_completions/mean_length": 121.49138259887695, "eval_completions/mean_terminated_length": 121.49138259887695, "eval_completions/min_length": 62.5, "eval_completions/min_terminated_length": 62.5, "eval_loss": 0.0, "eval_num_tokens": 172870400.0, "eval_reward": 1.028822124004364, "eval_reward_std": 0.16236025094985962, "eval_rewards/accuracy_reward": 0.26171875, "eval_rewards/brier_reward": 0.7987500131130219, "eval_rewards/confidence_one_or_zero": 0.033203125, "eval_rewards/format_reward": 1.0, "eval_rewards/mean_confidence_reward": 0.2767578214406967, "eval_rewards/smce": -0.028251958079636097, "eval_runtime": 10.2637, "eval_samples_per_second": 48.715, "eval_steps_per_second": 0.195, "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": 0.0005859375, "completions/max_length": 1320.6, "completions/max_terminated_length": 557.4, "completions/mean_length": 124.48154296875, "completions/mean_terminated_length": 123.65345611572266, "completions/min_length": 44.4, "completions/min_terminated_length": 44.4, "epoch": 0.176, "grad_norm": 0.0009458002750761807, "learning_rate": 1e-06, "loss": 0.0008, "num_tokens": 189382211.0, "reward": 1.0453623294830323, "reward_std": 0.10711575001478195, "rewards/accuracy_reward": 0.31328125, "rewards/brier_reward": 0.7821379065513611, "rewards/confidence_one_or_zero": 0.03759765625, "rewards/format_reward": 0.99853515625, "rewards/mean_confidence_reward": 0.29168972969055174, "rewards/smce": -0.03230265900492668, "step": 55 }, { "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": 0.0001953125, "completions/max_length": 885.6, "completions/max_terminated_length": 449.0, "completions/mean_length": 127.11484375, "completions/mean_terminated_length": 126.83966674804688, "completions/min_length": 58.6, "completions/min_terminated_length": 58.6, "epoch": 0.192, "grad_norm": 0.0010142099345102906, "learning_rate": 1e-06, "loss": 0.0006, "num_tokens": 205498683.0, "reward": 1.0576444864273071, "reward_std": 0.12386192083358764, "rewards/accuracy_reward": 0.34736328125, "rewards/brier_reward": 0.7734477996826172, "rewards/confidence_one_or_zero": 0.01337890625, "rewards/format_reward": 0.9990234375, "rewards/mean_confidence_reward": 0.34815791845321653, "rewards/smce": -0.04546214714646339, "step": 60 }, { "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": 0.00048828125, "completions/max_length": 1063.4, "completions/max_terminated_length": 436.0, "completions/mean_length": 134.46962890625, "completions/mean_terminated_length": 133.785693359375, "completions/min_length": 55.0, "completions/min_terminated_length": 55.0, "epoch": 0.208, "grad_norm": 0.0010471841087564826, "learning_rate": 1e-06, "loss": 0.0011, "num_tokens": 221907876.0, "reward": 1.0676575660705567, "reward_std": 0.13774634152650833, "rewards/accuracy_reward": 0.36845703125, "rewards/brier_reward": 0.7705541491508484, "rewards/confidence_one_or_zero": 0.0060546875, "rewards/format_reward": 0.9990234375, "rewards/mean_confidence_reward": 0.3837132155895233, "rewards/smce": -0.027201873436570167, "step": 65 }, { "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": 0.00029296875, "completions/max_length": 835.4, "completions/max_terminated_length": 387.2, "completions/mean_length": 131.84931640625, "completions/mean_terminated_length": 131.43808288574218, "completions/min_length": 58.4, "completions/min_terminated_length": 58.4, "epoch": 0.224, "grad_norm": 0.0010664016008377075, "learning_rate": 1e-06, "loss": 0.0014, "num_tokens": 238411197.0, "reward": 1.0667345762252807, "reward_std": 0.12648420929908752, "rewards/accuracy_reward": 0.3685546875, "rewards/brier_reward": 0.7694954991340637, "rewards/confidence_one_or_zero": 0.004296875, "rewards/format_reward": 0.99892578125, "rewards/mean_confidence_reward": 0.38269283771514895, "rewards/smce": -0.035074672847986224, "step": 70 }, { "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": 0.0001953125, "completions/max_length": 882.8, "completions/max_terminated_length": 421.2, "completions/mean_length": 131.6205078125, "completions/mean_terminated_length": 131.34631958007813, "completions/min_length": 58.4, "completions/min_terminated_length": 58.4, "epoch": 0.24, "grad_norm": 0.001074860105291009, "learning_rate": 1e-06, "loss": 0.0008, "num_tokens": 255010671.0, "reward": 1.0831371784210204, "reward_std": 0.12712678611278533, "rewards/accuracy_reward": 0.4126953125, "rewards/brier_reward": 0.7609473943710328, "rewards/confidence_one_or_zero": 0.0015625, "rewards/format_reward": 0.99921875, "rewards/mean_confidence_reward": 0.39454362392425535, "rewards/smce": -0.06587964631617069, "step": 75 }, { "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": 9.765625e-05, "completions/max_length": 596.8, "completions/max_terminated_length": 363.6, "completions/mean_length": 130.2615234375, "completions/mean_terminated_length": 130.1241485595703, "completions/min_length": 57.2, "completions/min_terminated_length": 57.2, "epoch": 0.256, "grad_norm": 0.0011320032645016909, "learning_rate": 1e-06, "loss": 0.0003, "num_tokens": 271399365.0, "reward": 1.0704259634017945, "reward_std": 0.13221554309129716, "rewards/accuracy_reward": 0.3767578125, "rewards/brier_reward": 0.7702070951461792, "rewards/confidence_one_or_zero": 0.00048828125, "rewards/format_reward": 0.99951171875, "rewards/mean_confidence_reward": 0.40313740968704226, "rewards/smce": -0.05625563189387321, "step": 80 }, { "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": 9.765625e-05, "completions/max_length": 683.0, "completions/max_terminated_length": 446.2, "completions/mean_length": 133.1107421875, "completions/mean_terminated_length": 132.97392883300782, "completions/min_length": 58.4, "completions/min_terminated_length": 58.4, "epoch": 0.272, "grad_norm": 0.0009479325381107628, "learning_rate": 1e-06, "loss": -0.0, "num_tokens": 287728115.0, "reward": 1.0818254709243775, "reward_std": 0.1228899359703064, "rewards/accuracy_reward": 0.40283203125, "rewards/brier_reward": 0.7644219875335694, "rewards/confidence_one_or_zero": 0.00048828125, "rewards/format_reward": 0.9998046875, "rewards/mean_confidence_reward": 0.39888736605644226, "rewards/smce": -0.03408620785921812, "step": 85 }, { "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": 0.0005859375, "completions/max_length": 1113.6, "completions/max_terminated_length": 411.6, "completions/mean_length": 134.308984375, "completions/mean_terminated_length": 133.48701171875, "completions/min_length": 59.4, "completions/min_terminated_length": 59.4, "epoch": 0.288, "grad_norm": 0.0009760709363035858, "learning_rate": 1e-06, "loss": 0.0014, "num_tokens": 304061615.0, "reward": 1.0798555850982665, "reward_std": 0.1159628912806511, "rewards/accuracy_reward": 0.39912109375, "rewards/brier_reward": 0.766473650932312, "rewards/confidence_one_or_zero": 0.000390625, "rewards/format_reward": 0.99912109375, "rewards/mean_confidence_reward": 0.3930133879184723, "rewards/smce": -0.050054707750678065, "step": 90 }, { "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": 0.0001953125, "completions/max_length": 866.4, "completions/max_terminated_length": 438.0, "completions/mean_length": 136.6109375, "completions/mean_terminated_length": 136.33718872070312, "completions/min_length": 64.4, "completions/min_terminated_length": 64.4, "epoch": 0.304, "grad_norm": 0.0010910023702308536, "learning_rate": 1e-06, "loss": 0.0005, "num_tokens": 320390463.0, "reward": 1.0854009866714478, "reward_std": 0.11015888154506684, "rewards/accuracy_reward": 0.41083984375, "rewards/brier_reward": 0.7642677545547485, "rewards/confidence_one_or_zero": 9.765625e-05, "rewards/format_reward": 0.99970703125, "rewards/mean_confidence_reward": 0.39040674567222594, "rewards/smce": -0.0401339016854763, "step": 95 }, { "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": 9.765625e-05, "completions/max_length": 650.6, "completions/max_terminated_length": 416.2, "completions/mean_length": 141.7556640625, "completions/mean_terminated_length": 141.61908874511718, "completions/min_length": 61.0, "completions/min_terminated_length": 61.0, "epoch": 0.32, "grad_norm": 0.0009000017307698727, "learning_rate": 1e-06, "loss": 0.0002, "num_tokens": 336930745.0, "reward": 1.0933893203735352, "reward_std": 0.10807573348283768, "rewards/accuracy_reward": 0.42783203125, "rewards/brier_reward": 0.7655725359916687, "rewards/confidence_one_or_zero": 0.00029296875, "rewards/format_reward": 0.99970703125, "rewards/mean_confidence_reward": 0.4067640006542206, "rewards/smce": -0.06333717741072178, "step": 100 }, { "epoch": 0.32, "eval_completions/clipped_ratio": 0.0, "eval_completions/max_length": 281.5, "eval_completions/max_terminated_length": 281.5, "eval_completions/mean_length": 144.8693618774414, "eval_completions/mean_terminated_length": 144.8693618774414, "eval_completions/min_length": 73.0, "eval_completions/min_terminated_length": 73.0, "eval_loss": 0.0, "eval_num_tokens": 336930745.0, "eval_reward": 1.0506283044815063, "eval_reward_std": 0.2293572798371315, "eval_rewards/accuracy_reward": 0.32421875, "eval_rewards/brier_reward": 0.7854808866977692, "eval_rewards/confidence_one_or_zero": 0.0, "eval_rewards/format_reward": 1.0, "eval_rewards/mean_confidence_reward": 0.4080800861120224, "eval_rewards/smce": -0.08443867601454258, "eval_runtime": 10.6049, "eval_samples_per_second": 47.148, "eval_steps_per_second": 0.189, "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": 0.00029296875, "completions/max_length": 881.4, "completions/max_terminated_length": 410.6, "completions/mean_length": 146.67197265625, "completions/mean_terminated_length": 146.263525390625, "completions/min_length": 63.2, "completions/min_terminated_length": 63.2, "epoch": 0.336, "grad_norm": 0.000741802214179188, "learning_rate": 1e-06, "loss": 0.0007, "num_tokens": 353155098.0, "reward": 1.0966195583343505, "reward_std": 0.10717584788799286, "rewards/accuracy_reward": 0.43037109375, "rewards/brier_reward": 0.7681578040122986, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.999609375, "rewards/mean_confidence_reward": 0.41693392395973206, "rewards/smce": -0.04899918362498283, "step": 105 }, { "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": 412.8, "completions/max_terminated_length": 412.8, "completions/mean_length": 152.62900390625, "completions/mean_terminated_length": 152.62900390625, "completions/min_length": 61.6, "completions/min_terminated_length": 61.6, "epoch": 0.352, "grad_norm": 0.0008869222365319729, "learning_rate": 1e-06, "loss": 0.0008, "num_tokens": 369978435.0, "reward": 1.0743722677230836, "reward_std": 0.12426654249429703, "rewards/accuracy_reward": 0.380078125, "rewards/brier_reward": 0.7729269146919251, "rewards/confidence_one_or_zero": 9.765625e-05, "rewards/format_reward": 0.99970703125, "rewards/mean_confidence_reward": 0.4185358822345734, "rewards/smce": -0.03968314975500107, "step": 110 }, { "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": 0.0001953125, "completions/max_length": 874.8, "completions/max_terminated_length": 451.2, "completions/mean_length": 153.2439453125, "completions/mean_terminated_length": 152.97373657226564, "completions/min_length": 59.6, "completions/min_terminated_length": 59.6, "epoch": 0.368, "grad_norm": 0.000924785912502557, "learning_rate": 1e-06, "loss": 0.0009, "num_tokens": 386613125.0, "reward": 1.0886813402175903, "reward_std": 0.11303802132606507, "rewards/accuracy_reward": 0.40703125, "rewards/brier_reward": 0.7744149088859558, "rewards/confidence_one_or_zero": 9.765625e-05, "rewards/format_reward": 0.99951171875, "rewards/mean_confidence_reward": 0.42386980056762696, "rewards/smce": -0.03596003744751215, "step": 115 }, { "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": 361.6, "completions/max_terminated_length": 361.6, "completions/mean_length": 149.928125, "completions/mean_terminated_length": 149.928125, "completions/min_length": 68.0, "completions/min_terminated_length": 68.0, "epoch": 0.384, "grad_norm": 0.0007946203113533556, "learning_rate": 1e-06, "loss": 0.0001, "num_tokens": 403004901.0, "reward": 1.0991857767105102, "reward_std": 0.10030071586370468, "rewards/accuracy_reward": 0.42626953125, "rewards/brier_reward": 0.7759941816329956, "rewards/confidence_one_or_zero": 9.765625e-05, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.41403782963752744, "rewards/smce": -0.037953479960560796, "step": 120 }, { "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": 9.765625e-05, "completions/max_length": 652.4, "completions/max_terminated_length": 504.6, "completions/mean_length": 149.6595703125, "completions/mean_terminated_length": 149.52425842285157, "completions/min_length": 64.2, "completions/min_terminated_length": 64.2, "epoch": 0.4, "grad_norm": 0.0008367216214537621, "learning_rate": 1e-06, "loss": 0.0008, "num_tokens": 419573863.0, "reward": 1.0881041288375854, "reward_std": 0.10986830145120621, "rewards/accuracy_reward": 0.40126953125, "rewards/brier_reward": 0.7783864617347718, "rewards/confidence_one_or_zero": 0.000390625, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.41706185340881347, "rewards/smce": -0.03350854460150003, "step": 125 }, { "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": 441.4, "completions/max_terminated_length": 441.4, "completions/mean_length": 151.35322265625, "completions/mean_terminated_length": 151.35322265625, "completions/min_length": 70.0, "completions/min_terminated_length": 70.0, "epoch": 0.416, "grad_norm": 0.0007317252457141876, "learning_rate": 1e-06, "loss": 0.0002, "num_tokens": 436004904.0, "reward": 1.0960778713226318, "reward_std": 0.106614151597023, "rewards/accuracy_reward": 0.41640625, "rewards/brier_reward": 0.7772732257843018, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 1.0, "rewards/mean_confidence_reward": 0.41116663813591003, "rewards/smce": -0.015245692990720271, "step": 130 }, { "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": 9.765625e-05, "completions/max_length": 654.4, "completions/max_terminated_length": 417.6, "completions/mean_length": 154.1056640625, "completions/mean_terminated_length": 153.97059631347656, "completions/min_length": 72.0, "completions/min_terminated_length": 72.0, "epoch": 0.432, "grad_norm": 0.0007601842517033219, "learning_rate": 1e-06, "loss": 0.0001, "num_tokens": 452597282.0, "reward": 1.1045128583908081, "reward_std": 0.10084618777036666, "rewards/accuracy_reward": 0.43818359375, "rewards/brier_reward": 0.7752175211906434, "rewards/confidence_one_or_zero": 9.765625e-05, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.4205751955509186, "rewards/smce": -0.04278548639267683, "step": 135 }, { "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": 435.2, "completions/max_terminated_length": 435.2, "completions/mean_length": 157.68974609375, "completions/mean_terminated_length": 157.68974609375, "completions/min_length": 72.8, "completions/min_terminated_length": 72.8, "epoch": 0.448, "grad_norm": 0.0006946580251678824, "learning_rate": 1e-06, "loss": 0.0, "num_tokens": 469164825.0, "reward": 1.1021424531936646, "reward_std": 0.10896257162094117, "rewards/accuracy_reward": 0.4326171875, "rewards/brier_reward": 0.7751887559890747, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.4384663462638855, "rewards/smce": -0.03424236755818129, "step": 140 }, { "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": 0.0001953125, "completions/max_length": 879.2, "completions/max_terminated_length": 420.6, "completions/mean_length": 159.316796875, "completions/mean_terminated_length": 159.04764404296876, "completions/min_length": 72.6, "completions/min_terminated_length": 72.6, "epoch": 0.464, "grad_norm": 0.0008301216876134276, "learning_rate": 1e-06, "loss": 0.0005, "num_tokens": 485967045.0, "reward": 1.076231551170349, "reward_std": 0.11073880642652512, "rewards/accuracy_reward": 0.3748046875, "rewards/brier_reward": 0.7852112293243408, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.9998046875, "rewards/mean_confidence_reward": 0.4326315402984619, "rewards/smce": -0.07358203530311584, "step": 145 }, { "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": 9.765625e-05, "completions/max_length": 629.8, "completions/max_terminated_length": 417.4, "completions/mean_length": 160.202734375, "completions/mean_terminated_length": 160.06853942871095, "completions/min_length": 74.2, "completions/min_terminated_length": 74.2, "epoch": 0.48, "grad_norm": 0.0008279711473733187, "learning_rate": 1e-06, "loss": 0.0007, "num_tokens": 502655553.0, "reward": 1.106163740158081, "reward_std": 0.10428445488214493, "rewards/accuracy_reward": 0.4455078125, "rewards/brier_reward": 0.7723424196243286, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.9998046875, "rewards/mean_confidence_reward": 0.41151692271232604, "rewards/smce": -0.053282588720321655, "step": 150 }, { "epoch": 0.48, "eval_completions/clipped_ratio": 0.0, "eval_completions/max_length": 348.5, "eval_completions/max_terminated_length": 348.5, "eval_completions/mean_length": 159.40221405029297, "eval_completions/mean_terminated_length": 159.40221405029297, "eval_completions/min_length": 76.5, "eval_completions/min_terminated_length": 76.5, "eval_loss": 0.0, "eval_num_tokens": 502655553.0, "eval_reward": 1.0689831376075745, "eval_reward_std": 0.23300867527723312, "eval_rewards/accuracy_reward": 0.345703125, "eval_rewards/brier_reward": 0.7980294525623322, "eval_rewards/confidence_one_or_zero": 0.0, "eval_rewards/format_reward": 1.0, "eval_rewards/mean_confidence_reward": 0.4022851586341858, "eval_rewards/smce": -0.05767090246081352, "eval_runtime": 11.9118, "eval_samples_per_second": 41.975, "eval_steps_per_second": 0.168, "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": 0.0, "completions/max_length": 404.0, "completions/max_terminated_length": 404.0, "completions/mean_length": 157.42607421875, "completions/mean_terminated_length": 157.42607421875, "completions/min_length": 72.0, "completions/min_terminated_length": 72.0, "epoch": 0.496, "grad_norm": 0.0008031203760765493, "learning_rate": 1e-06, "loss": 0.0002, "num_tokens": 519575436.0, "reward": 1.1064825534820557, "reward_std": 0.10670529305934906, "rewards/accuracy_reward": 0.44521484375, "rewards/brier_reward": 0.7713125824928284, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 1.0, "rewards/mean_confidence_reward": 0.42044044137001035, "rewards/smce": -0.03563190307468176, "step": 155 }, { "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": 376.8, "completions/max_terminated_length": 376.8, "completions/mean_length": 156.19970703125, "completions/mean_terminated_length": 156.19970703125, "completions/min_length": 78.0, "completions/min_terminated_length": 78.0, "epoch": 0.512, "grad_norm": 0.0009704737458378077, "learning_rate": 1e-06, "loss": -0.0004, "num_tokens": 536320585.0, "reward": 1.1233646392822265, "reward_std": 0.10814391821622849, "rewards/accuracy_reward": 0.47216796875, "rewards/brier_reward": 0.7789788365364074, "rewards/confidence_one_or_zero": 9.765625e-05, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.4404987394809723, "rewards/smce": -0.04320650082081556, "step": 160 }, { "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": 0.0001953125, "completions/max_length": 869.6, "completions/max_terminated_length": 417.2, "completions/mean_length": 160.86787109375, "completions/mean_terminated_length": 160.59920349121094, "completions/min_length": 84.0, "completions/min_terminated_length": 84.0, "epoch": 0.528, "grad_norm": 0.0009864222956821322, "learning_rate": 1e-06, "loss": 0.0007, "num_tokens": 552997408.0, "reward": 1.1117675065994264, "reward_std": 0.12492350935935974, "rewards/accuracy_reward": 0.4421875, "rewards/brier_reward": 0.7852141976356506, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.99970703125, "rewards/mean_confidence_reward": 0.47042285203933715, "rewards/smce": -0.035746574960649015, "step": 165 }, { "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": 0.0001953125, "completions/max_length": 883.6, "completions/max_terminated_length": 435.2, "completions/mean_length": 159.598828125, "completions/mean_terminated_length": 159.33026428222655, "completions/min_length": 80.4, "completions/min_terminated_length": 80.4, "epoch": 0.544, "grad_norm": 0.0008369262213818729, "learning_rate": 1e-06, "loss": 0.0009, "num_tokens": 569795284.0, "reward": 1.1314317226409911, "reward_std": 0.11883504688739777, "rewards/accuracy_reward": 0.49619140625, "rewards/brier_reward": 0.7696640849113464, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.9998046875, "rewards/mean_confidence_reward": 0.48162569403648375, "rewards/smce": -0.027976423129439355, "step": 170 }, { "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": 354.2, "completions/max_terminated_length": 354.2, "completions/mean_length": 153.4052734375, "completions/mean_terminated_length": 153.4052734375, "completions/min_length": 73.0, "completions/min_terminated_length": 73.0, "epoch": 0.56, "grad_norm": 0.0009207469411194324, "learning_rate": 1e-06, "loss": 0.0003, "num_tokens": 586187562.0, "reward": 1.099264669418335, "reward_std": 0.12063608765602112, "rewards/accuracy_reward": 0.41640625, "rewards/brier_reward": 0.7878692507743835, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.4717119216918945, "rewards/smce": -0.056493572890758514, "step": 175 }, { "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": 0.0001953125, "completions/max_length": 830.0, "completions/max_terminated_length": 334.2, "completions/mean_length": 150.38310546875, "completions/mean_terminated_length": 150.1125, "completions/min_length": 72.2, "completions/min_terminated_length": 72.2, "epoch": 0.576, "grad_norm": 0.000673173344694078, "learning_rate": 1e-06, "loss": 0.0005, "num_tokens": 602914109.0, "reward": 1.10039484500885, "reward_std": 0.09124593585729598, "rewards/accuracy_reward": 0.42099609375, "rewards/brier_reward": 0.7848548769950867, "rewards/confidence_one_or_zero": 0.0001953125, "rewards/format_reward": 0.9998046875, "rewards/mean_confidence_reward": 0.4232636749744415, "rewards/smce": -0.04866776242852211, "step": 180 }, { "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": 390.8, "completions/max_terminated_length": 390.8, "completions/mean_length": 148.59873046875, "completions/mean_terminated_length": 148.59873046875, "completions/min_length": 74.2, "completions/min_terminated_length": 74.2, "epoch": 0.592, "grad_norm": 0.0007362031610682607, "learning_rate": 1e-06, "loss": -0.0001, "num_tokens": 619603472.0, "reward": 1.109256672859192, "reward_std": 0.08407505601644516, "rewards/accuracy_reward": 0.4404296875, "rewards/brier_reward": 0.7812385320663452, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 1.0, "rewards/mean_confidence_reward": 0.41702051758766173, "rewards/smce": -0.03155655637383461, "step": 185 }, { "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": 427.6, "completions/max_terminated_length": 427.6, "completions/mean_length": 152.87138671875, "completions/mean_terminated_length": 152.87138671875, "completions/min_length": 78.6, "completions/min_terminated_length": 78.6, "epoch": 0.608, "grad_norm": 0.0009035294642671943, "learning_rate": 1e-06, "loss": 0.0001, "num_tokens": 636168363.0, "reward": 1.1182846546173095, "reward_std": 0.09603894054889679, "rewards/accuracy_reward": 0.44208984375, "rewards/brier_reward": 0.7972625851631164, "rewards/confidence_one_or_zero": 9.765625e-05, "rewards/format_reward": 1.0, "rewards/mean_confidence_reward": 0.43902148604393004, "rewards/smce": -0.02784018460661173, "step": 190 }, { "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": 0.0001953125, "completions/max_length": 896.8, "completions/max_terminated_length": 418.2, "completions/mean_length": 159.56494140625, "completions/mean_terminated_length": 159.29615173339843, "completions/min_length": 79.8, "completions/min_terminated_length": 79.8, "epoch": 0.624, "grad_norm": 0.0011967478785663843, "learning_rate": 1e-06, "loss": 0.0009, "num_tokens": 653146212.0, "reward": 1.1095570087432862, "reward_std": 0.11243704110383987, "rewards/accuracy_reward": 0.43642578125, "rewards/brier_reward": 0.7865937709808349, "rewards/confidence_one_or_zero": 9.765625e-05, "rewards/format_reward": 0.9998046875, "rewards/mean_confidence_reward": 0.46078904867172243, "rewards/smce": -0.037111565470695496, "step": 195 }, { "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": 474.2, "completions/max_terminated_length": 474.2, "completions/mean_length": 163.27529296875, "completions/mean_terminated_length": 163.27529296875, "completions/min_length": 80.4, "completions/min_terminated_length": 80.4, "epoch": 0.64, "grad_norm": 0.0007251665811054409, "learning_rate": 1e-06, "loss": 0.0004, "num_tokens": 670160839.0, "reward": 1.1448447942733764, "reward_std": 0.09686506688594818, "rewards/accuracy_reward": 0.5076171875, "rewards/brier_reward": 0.7882843375205993, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 1.0, "rewards/mean_confidence_reward": 0.466916024684906, "rewards/smce": -0.06212771981954575, "step": 200 }, { "epoch": 0.64, "eval_completions/clipped_ratio": 0.0, "eval_completions/max_length": 316.5, "eval_completions/max_terminated_length": 316.5, "eval_completions/mean_length": 167.25476837158203, "eval_completions/mean_terminated_length": 167.25476837158203, "eval_completions/min_length": 88.0, "eval_completions/min_terminated_length": 88.0, "eval_loss": 0.0, "eval_num_tokens": 670160839.0, "eval_reward": 1.0876256227493286, "eval_reward_std": 0.26613427698612213, "eval_rewards/accuracy_reward": 0.384765625, "eval_rewards/brier_reward": 0.7975634634494781, "eval_rewards/confidence_one_or_zero": 0.0, "eval_rewards/format_reward": 1.0, "eval_rewards/mean_confidence_reward": 0.45537109673023224, "eval_rewards/smce": -0.0707861389964819, "eval_runtime": 10.9563, "eval_samples_per_second": 45.636, "eval_steps_per_second": 0.183, "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": 9.765625e-05, "completions/max_length": 671.4, "completions/max_terminated_length": 449.6, "completions/mean_length": 164.71435546875, "completions/mean_terminated_length": 164.58023071289062, "completions/min_length": 81.2, "completions/min_terminated_length": 81.2, "epoch": 0.656, "grad_norm": 0.0010223239660263062, "learning_rate": 1e-06, "loss": 0.0, "num_tokens": 686704058.0, "reward": 1.0958014249801635, "reward_std": 0.11311615407466888, "rewards/accuracy_reward": 0.41611328125, "rewards/brier_reward": 0.7816610336303711, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.47294239401817323, "rewards/smce": -0.060746995359659196, "step": 205 }, { "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": 437.4, "completions/max_terminated_length": 437.4, "completions/mean_length": 169.5923828125, "completions/mean_terminated_length": 169.5923828125, "completions/min_length": 86.2, "completions/min_terminated_length": 86.2, "epoch": 0.672, "grad_norm": 0.0006438858108595014, "learning_rate": 1e-06, "loss": 0.0001, "num_tokens": 703354124.0, "reward": 1.1204289674758912, "reward_std": 0.08563904613256454, "rewards/accuracy_reward": 0.45576171875, "rewards/brier_reward": 0.7892528295516967, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 1.0, "rewards/mean_confidence_reward": 0.42501270174980166, "rewards/smce": -0.04157466329634189, "step": 210 }, { "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": 9.765625e-05, "completions/max_length": 689.0, "completions/max_terminated_length": 480.4, "completions/mean_length": 176.6537109375, "completions/mean_terminated_length": 176.5207946777344, "completions/min_length": 87.2, "completions/min_terminated_length": 87.2, "epoch": 0.688, "grad_norm": 0.0008102591382339597, "learning_rate": 1e-06, "loss": 0.0005, "num_tokens": 720116978.0, "reward": 1.1223333358764649, "reward_std": 0.09338834285736083, "rewards/accuracy_reward": 0.4533203125, "rewards/brier_reward": 0.7958316683769227, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.4228310644626617, "rewards/smce": -0.04388436172157526, "step": 215 }, { "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": 9.765625e-05, "completions/max_length": 685.0, "completions/max_terminated_length": 484.0, "completions/mean_length": 179.18369140625, "completions/mean_terminated_length": 179.05135498046874, "completions/min_length": 93.6, "completions/min_terminated_length": 93.6, "epoch": 0.704, "grad_norm": 0.0007676195236854255, "learning_rate": 1e-06, "loss": 0.0005, "num_tokens": 736817963.0, "reward": 1.131119704246521, "reward_std": 0.0943239152431488, "rewards/accuracy_reward": 0.47373046875, "rewards/brier_reward": 0.7917332649230957, "rewards/confidence_one_or_zero": 9.765625e-05, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.44699220061302186, "rewards/smce": -0.03127504363656044, "step": 220 }, { "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": 9.765625e-05, "completions/max_length": 741.0, "completions/max_terminated_length": 503.2, "completions/mean_length": 180.8234375, "completions/mean_terminated_length": 180.69107055664062, "completions/min_length": 90.0, "completions/min_terminated_length": 90.0, "epoch": 0.72, "grad_norm": 0.0009789899922907352, "learning_rate": 1e-06, "loss": 0.0005, "num_tokens": 753679451.0, "reward": 1.1428563594818115, "reward_std": 0.10433020591735839, "rewards/accuracy_reward": 0.4857421875, "rewards/brier_reward": 0.8039116024971008, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.504841786623001, "rewards/smce": -0.0384433601051569, "step": 225 }, { "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": 483.6, "completions/max_terminated_length": 483.6, "completions/mean_length": 183.2181640625, "completions/mean_terminated_length": 183.2181640625, "completions/min_length": 85.2, "completions/min_terminated_length": 85.2, "epoch": 0.736, "grad_norm": 0.0009930457454174757, "learning_rate": 1e-06, "loss": 0.0, "num_tokens": 770495189.0, "reward": 1.1361095428466796, "reward_std": 0.123087178170681, "rewards/accuracy_reward": 0.49052734375, "rewards/brier_reward": 0.7868128776550293, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.5383701086044311, "rewards/smce": -0.05024341568350792, "step": 230 }, { "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": 9.765625e-05, "completions/max_length": 627.2, "completions/max_terminated_length": 419.0, "completions/mean_length": 185.18837890625, "completions/mean_terminated_length": 185.05672302246094, "completions/min_length": 87.8, "completions/min_terminated_length": 87.8, "epoch": 0.752, "grad_norm": 0.0009987555677071214, "learning_rate": 1e-06, "loss": 0.0002, "num_tokens": 787618718.0, "reward": 1.1308838367462157, "reward_std": 0.11320770680904388, "rewards/accuracy_reward": 0.47373046875, "rewards/brier_reward": 0.792686128616333, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.4944267630577087, "rewards/smce": -0.04552147686481476, "step": 235 }, { "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": 481.4, "completions/max_terminated_length": 481.4, "completions/mean_length": 191.2140625, "completions/mean_terminated_length": 191.2140625, "completions/min_length": 95.6, "completions/min_terminated_length": 95.6, "epoch": 0.768, "grad_norm": 0.0008614786202087998, "learning_rate": 1e-06, "loss": 0.0002, "num_tokens": 804509454.0, "reward": 1.136540699005127, "reward_std": 0.094244784116745, "rewards/accuracy_reward": 0.473828125, "rewards/brier_reward": 0.8010059595108032, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 1.0, "rewards/mean_confidence_reward": 0.47049121260643006, "rewards/smce": -0.017536511179059744, "step": 240 }, { "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": 9.765625e-05, "completions/max_length": 642.6, "completions/max_terminated_length": 434.0, "completions/mean_length": 192.61494140625, "completions/mean_terminated_length": 192.4836853027344, "completions/min_length": 101.0, "completions/min_terminated_length": 101.0, "epoch": 0.784, "grad_norm": 0.000979123986326158, "learning_rate": 1e-06, "loss": 0.0004, "num_tokens": 821656199.0, "reward": 1.1346184253692626, "reward_std": 0.09833287298679352, "rewards/accuracy_reward": 0.4875, "rewards/brier_reward": 0.7900451183319092, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.4683769643306732, "rewards/smce": -0.08211531937122345, "step": 245 }, { "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": 0.0001953125, "completions/max_length": 897.8, "completions/max_terminated_length": 464.0, "completions/mean_length": 194.98642578125, "completions/mean_terminated_length": 194.72418823242188, "completions/min_length": 98.2, "completions/min_terminated_length": 98.2, "epoch": 0.8, "grad_norm": 0.0008208116050809622, "learning_rate": 1e-06, "loss": 0.0006, "num_tokens": 838663420.0, "reward": 1.149678373336792, "reward_std": 0.10992196351289749, "rewards/accuracy_reward": 0.5005859375, "rewards/brier_reward": 0.8018845319747925, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.9998046875, "rewards/mean_confidence_reward": 0.4957675993442535, "rewards/smce": -0.029193690791726113, "step": 250 }, { "epoch": 0.8, "eval_completions/clipped_ratio": 0.0, "eval_completions/max_length": 396.5, "eval_completions/max_terminated_length": 396.5, "eval_completions/mean_length": 201.88136291503906, "eval_completions/mean_terminated_length": 201.88136291503906, "eval_completions/min_length": 116.5, "eval_completions/min_terminated_length": 116.5, "eval_loss": 0.0, "eval_num_tokens": 838663420.0, "eval_reward": 1.080160140991211, "eval_reward_std": 0.28166723251342773, "eval_rewards/accuracy_reward": 0.3671875, "eval_rewards/brier_reward": 0.8043310642242432, "eval_rewards/confidence_one_or_zero": 0.0, "eval_rewards/format_reward": 1.0, "eval_rewards/mean_confidence_reward": 0.47900390625, "eval_rewards/smce": -0.11199219152331352, "eval_runtime": 12.7377, "eval_samples_per_second": 39.253, "eval_steps_per_second": 0.157, "step": 250 }, { "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": 9.765625e-05, "completions/max_length": 752.4, "completions/max_terminated_length": 534.4, "completions/mean_length": 197.24560546875, "completions/mean_terminated_length": 197.114404296875, "completions/min_length": 102.2, "completions/min_terminated_length": 102.2, "epoch": 0.816, "grad_norm": 0.0008329146658070385, "learning_rate": 1e-06, "loss": 0.0007, "num_tokens": 855782383.0, "reward": 1.1449370622634887, "reward_std": 0.11253967136144638, "rewards/accuracy_reward": 0.48984375, "rewards/brier_reward": 0.802400529384613, "rewards/confidence_one_or_zero": 9.765625e-05, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.5037607312202453, "rewards/smce": -0.022733921930193902, "step": 255 }, { "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": 0.0001953125, "completions/max_length": 886.8, "completions/max_terminated_length": 460.2, "completions/mean_length": 204.29169921875, "completions/mean_terminated_length": 204.0310089111328, "completions/min_length": 104.4, "completions/min_terminated_length": 104.4, "epoch": 0.832, "grad_norm": 0.0008481323602609336, "learning_rate": 1e-06, "loss": 0.0007, "num_tokens": 872882682.0, "reward": 1.150204586982727, "reward_std": 0.09409180730581283, "rewards/accuracy_reward": 0.501171875, "rewards/brier_reward": 0.8049479007720948, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.9998046875, "rewards/mean_confidence_reward": 0.4569570302963257, "rewards/smce": -0.05516120456159115, "step": 260 }, { "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": 0.000390625, "completions/max_length": 1117.8, "completions/max_terminated_length": 500.0, "completions/mean_length": 202.97705078125, "completions/mean_terminated_length": 202.45654907226563, "completions/min_length": 105.0, "completions/min_terminated_length": 105.0, "epoch": 0.848, "grad_norm": 0.0008220033487305045, "learning_rate": 1e-06, "loss": 0.0011, "num_tokens": 889975535.0, "reward": 1.1360312700271606, "reward_std": 0.09862071722745895, "rewards/accuracy_reward": 0.4732421875, "rewards/brier_reward": 0.8020096898078919, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.999609375, "rewards/mean_confidence_reward": 0.4578730583190918, "rewards/smce": -0.027996178716421127, "step": 265 }, { "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": 0.0001953125, "completions/max_length": 975.6, "completions/max_terminated_length": 603.2, "completions/mean_length": 204.6900390625, "completions/mean_terminated_length": 204.43079223632813, "completions/min_length": 97.6, "completions/min_terminated_length": 97.6, "epoch": 0.864, "grad_norm": 0.0009190791170112789, "learning_rate": 1e-06, "loss": 0.0009, "num_tokens": 907058377.0, "reward": 1.154773187637329, "reward_std": 0.1105512484908104, "rewards/accuracy_reward": 0.512109375, "rewards/brier_reward": 0.8000603199005127, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.999609375, "rewards/mean_confidence_reward": 0.5104794800281525, "rewards/smce": -0.02233685217797756, "step": 270 }, { "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": 9.765625e-05, "completions/max_length": 801.8, "completions/max_terminated_length": 598.8, "completions/mean_length": 201.2875, "completions/mean_terminated_length": 201.1578857421875, "completions/min_length": 99.0, "completions/min_terminated_length": 99.0, "epoch": 0.88, "grad_norm": 0.0007379170856438577, "learning_rate": 1e-06, "loss": 0.0002, "num_tokens": 924266633.0, "reward": 1.1296437740325929, "reward_std": 0.10275594592094421, "rewards/accuracy_reward": 0.460546875, "rewards/brier_reward": 0.8046643137931824, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.9998046875, "rewards/mean_confidence_reward": 0.5026220738887787, "rewards/smce": -0.05729248225688934, "step": 275 }, { "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": 486.6, "completions/max_terminated_length": 486.6, "completions/mean_length": 199.25869140625, "completions/mean_terminated_length": 199.25869140625, "completions/min_length": 100.6, "completions/min_terminated_length": 100.6, "epoch": 0.896, "grad_norm": 0.0009705667616799474, "learning_rate": 1e-06, "loss": 0.0002, "num_tokens": 941417890.0, "reward": 1.1350660562515258, "reward_std": 0.09697622209787368, "rewards/accuracy_reward": 0.47099609375, "rewards/brier_reward": 0.804245901107788, "rewards/confidence_one_or_zero": 9.765625e-05, "rewards/format_reward": 0.9998046875, "rewards/mean_confidence_reward": 0.47399707436561583, "rewards/smce": -0.049154537916183474, "step": 280 }, { "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": 9.765625e-05, "completions/max_length": 692.0, "completions/max_terminated_length": 475.6, "completions/mean_length": 201.79638671875, "completions/mean_terminated_length": 201.66575927734374, "completions/min_length": 94.8, "completions/min_terminated_length": 94.8, "epoch": 0.912, "grad_norm": 0.0009069403749890625, "learning_rate": 1e-06, "loss": 0.0004, "num_tokens": 958535581.0, "reward": 1.1326754093170166, "reward_std": 0.09280429035425186, "rewards/accuracy_reward": 0.466796875, "rewards/brier_reward": 0.8028522849082946, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.99990234375, "rewards/mean_confidence_reward": 0.4540849685668945, "rewards/smce": -0.04201528280973434, "step": 285 }, { "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": 0.0001953125, "completions/max_length": 868.4, "completions/max_terminated_length": 445.2, "completions/mean_length": 200.70908203125, "completions/mean_terminated_length": 200.4482849121094, "completions/min_length": 102.4, "completions/min_terminated_length": 102.4, "epoch": 0.928, "grad_norm": 0.0009814233053475618, "learning_rate": 1e-06, "loss": 0.0008, "num_tokens": 975617658.0, "reward": 1.143230175971985, "reward_std": 0.09210614264011383, "rewards/accuracy_reward": 0.4984375, "rewards/brier_reward": 0.7931683301925659, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.9998046875, "rewards/mean_confidence_reward": 0.47999219298362733, "rewards/smce": -0.04950987063348293, "step": 290 }, { "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": 0.00029296875, "completions/max_length": 1091.2, "completions/max_terminated_length": 500.2, "completions/mean_length": 200.055859375, "completions/mean_terminated_length": 199.66469116210936, "completions/min_length": 98.0, "completions/min_terminated_length": 98.0, "epoch": 0.944, "grad_norm": 0.0007579100783914328, "learning_rate": 1e-06, "loss": 0.0003, "num_tokens": 992641654.0, "reward": 1.1395776748657227, "reward_std": 0.1035296618938446, "rewards/accuracy_reward": 0.4876953125, "rewards/brier_reward": 0.7984759330749511, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.99970703125, "rewards/mean_confidence_reward": 0.4942158281803131, "rewards/smce": -0.06723861880600453, "step": 295 }, { "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": 0.000390625, "completions/max_length": 755.6, "completions/max_terminated_length": 591.6, "completions/mean_length": 203.0103515625, "completions/mean_terminated_length": 202.48793029785156, "completions/min_length": 103.2, "completions/min_terminated_length": 103.2, "epoch": 0.96, "grad_norm": 0.0008992342045530677, "learning_rate": 1e-06, "loss": 0.0004, "num_tokens": 1009660800.0, "reward": 1.1296122550964356, "reward_std": 0.09960427284240722, "rewards/accuracy_reward": 0.46044921875, "rewards/brier_reward": 0.8044404029846192, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.999609375, "rewards/mean_confidence_reward": 0.5082920014858245, "rewards/smce": -0.05275399778038263, "step": 300 }, { "epoch": 0.96, "eval_completions/clipped_ratio": 0.0, "eval_completions/max_length": 347.5, "eval_completions/max_terminated_length": 347.5, "eval_completions/mean_length": 202.71531677246094, "eval_completions/mean_terminated_length": 202.71531677246094, "eval_completions/min_length": 107.0, "eval_completions/min_terminated_length": 107.0, "eval_loss": 0.0, "eval_num_tokens": 1009660800.0, "eval_reward": 1.1001547574996948, "eval_reward_std": 0.2800827920436859, "eval_rewards/accuracy_reward": 0.39453125, "eval_rewards/brier_reward": 0.8129736185073853, "eval_rewards/confidence_one_or_zero": 0.0, "eval_rewards/format_reward": 1.0, "eval_rewards/mean_confidence_reward": 0.4657226651906967, "eval_rewards/smce": -0.07196288928389549, "eval_runtime": 12.214, "eval_samples_per_second": 40.937, "eval_steps_per_second": 0.164, "step": 300 }, { "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": 0.00048828125, "completions/max_length": 902.2, "completions/max_terminated_length": 543.6, "completions/mean_length": 201.16455078125, "completions/mean_terminated_length": 200.51107788085938, "completions/min_length": 103.2, "completions/min_terminated_length": 103.2, "epoch": 0.976, "grad_norm": 0.0007909019477665424, "learning_rate": 1e-06, "loss": 0.0014, "num_tokens": 1026581845.0, "reward": 1.1536278963088988, "reward_std": 0.0806884527206421, "rewards/accuracy_reward": 0.5001953125, "rewards/brier_reward": 0.8116249918937684, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 0.99951171875, "rewards/mean_confidence_reward": 0.46386427283287046, "rewards/smce": -0.04077114947140217, "step": 305 }, { "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": 559.6, "completions/max_terminated_length": 559.6, "completions/mean_length": 195.6953125, "completions/mean_terminated_length": 195.6953125, "completions/min_length": 102.2, "completions/min_terminated_length": 102.2, "epoch": 0.992, "grad_norm": 0.0007479905034415424, "learning_rate": 1e-06, "loss": 0.0009, "num_tokens": 1043714245.0, "reward": 1.1366238832473754, "reward_std": 0.09564611464738845, "rewards/accuracy_reward": 0.480859375, "rewards/brier_reward": 0.7965283155441284, "rewards/confidence_one_or_zero": 0.0001953125, "rewards/format_reward": 1.0, "rewards/mean_confidence_reward": 0.4687353610992432, "rewards/smce": -0.04140814319252968, "step": 310 }, { "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": 442.0, "completions/max_terminated_length": 442.0, "completions/mean_length": 192.87665557861328, "completions/mean_terminated_length": 192.87665557861328, "completions/min_length": 100.5, "completions/min_terminated_length": 100.5, "epoch": 0.9984, "num_tokens": 1050511528.0, "reward": 1.135094702243805, "reward_std": 0.10495336353778839, "rewards/accuracy_reward": 0.48291015625, "rewards/brier_reward": 0.7914325296878815, "rewards/confidence_one_or_zero": 0.0, "rewards/format_reward": 1.0, "rewards/mean_confidence_reward": 0.5201000869274139, "rewards/smce": -0.04154240898787975, "step": 312, "total_flos": 0.0, "train_loss": 0.00407086915840456, "train_runtime": 42825.4883, "train_samples_per_second": 0.467, "train_steps_per_second": 0.007 } ], "logging_steps": 5, "max_steps": 312, "num_input_tokens_seen": 1050511528, "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 }