trainer_state.json

{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.995499549954996,
  "eval_steps": 15,
  "global_step": 395,
  "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": 0.030133928571428603,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2017.0,
      "completions/mean_length": 559.7366333007812,
      "completions/mean_terminated_length": 513.4959716796875,
      "completions/min_length": 4.0,
      "completions/min_terminated_length": 4.0,
      "epoch": 0.012601260126012601,
      "grad_norm": 0.40923070907592773,
      "learning_rate": 0.0,
      "loss": 0.0084,
      "num_tokens": 629108.0,
      "reward": 0.263950914144516,
      "reward_std": 0.3971380889415741,
      "rewards/accuracy_reward/mean": 0.1607142835855484,
      "rewards/accuracy_reward/std": 0.3674723207950592,
      "rewards/cosine_scaled_reward/mean": 0.1607142835855484,
      "rewards/cosine_scaled_reward/std": 0.3674723207950592,
      "rewards/format_reward/mean": 0.0457589291036129,
      "rewards/format_reward/std": 0.20907855033874512,
      "step": 1
    },
    {
      "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": 0.01981026785714285,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1937.5,
      "completions/mean_length": 523.0039215087891,
      "completions/mean_terminated_length": 492.1688919067383,
      "completions/min_length": 3.5,
      "completions/min_terminated_length": 3.5,
      "epoch": 0.063006300630063,
      "grad_norm": 0.43882623314857483,
      "learning_rate": 1e-07,
      "loss": -0.0058,
      "num_tokens": 3051026.0,
      "reward": 0.25034814700484276,
      "reward_std": 0.38232600688934326,
      "rewards/accuracy_reward/mean": 0.1526227667927742,
      "rewards/accuracy_reward/std": 0.35816748440265656,
      "rewards/cosine_scaled_reward/mean": 0.15248198807239532,
      "rewards/cosine_scaled_reward/std": 0.3578503802418709,
      "rewards/format_reward/mean": 0.04296875,
      "rewards/format_reward/std": 0.2016594335436821,
      "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": 0.02522321428571428,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2008.8,
      "completions/mean_length": 549.8663208007813,
      "completions/mean_terminated_length": 511.09259643554685,
      "completions/min_length": 4.8,
      "completions/min_terminated_length": 4.8,
      "epoch": 0.126012601260126,
      "grad_norm": 0.3167170286178589,
      "learning_rate": 2.25e-07,
      "loss": -0.0017,
      "num_tokens": 6199283.0,
      "reward": 0.2494979292154312,
      "reward_std": 0.3860060155391693,
      "rewards/accuracy_reward/mean": 0.15089285969734192,
      "rewards/accuracy_reward/std": 0.3568773686885834,
      "rewards/cosine_scaled_reward/mean": 0.1507815644145012,
      "rewards/cosine_scaled_reward/std": 0.3566217184066772,
      "rewards/format_reward/mean": 0.04642857164144516,
      "rewards/format_reward/std": 0.21049095690250397,
      "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": 0.022767857142857163,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1946.6,
      "completions/mean_length": 550.6145263671875,
      "completions/mean_terminated_length": 515.6882019042969,
      "completions/min_length": 8.6,
      "completions/min_terminated_length": 8.6,
      "epoch": 0.18901890189018902,
      "grad_norm": 0.34566730260849,
      "learning_rate": 3.5e-07,
      "loss": 0.0061,
      "num_tokens": 9310292.0,
      "reward": 0.2928151488304138,
      "reward_std": 0.40939453840255735,
      "rewards/accuracy_reward/mean": 0.17410714328289031,
      "rewards/accuracy_reward/std": 0.378982412815094,
      "rewards/cosine_scaled_reward/mean": 0.1740231305360794,
      "rewards/cosine_scaled_reward/std": 0.37881343960762026,
      "rewards/format_reward/mean": 0.0633928582072258,
      "rewards/format_reward/std": 0.24350899159908296,
      "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": 0.02901785714285714,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1996.8,
      "completions/mean_length": 563.8605102539062,
      "completions/mean_terminated_length": 519.6845397949219,
      "completions/min_length": 3.6,
      "completions/min_terminated_length": 3.6,
      "epoch": 0.252025202520252,
      "grad_norm": 6.714093208312988,
      "learning_rate": 4.7499999999999995e-07,
      "loss": -0.0035,
      "num_tokens": 12491619.0,
      "reward": 0.3050673246383667,
      "reward_std": 0.425398576259613,
      "rewards/accuracy_reward/mean": 0.16316964328289033,
      "rewards/accuracy_reward/std": 0.36818013787269593,
      "rewards/cosine_scaled_reward/mean": 0.16303642094135284,
      "rewards/cosine_scaled_reward/std": 0.3678862988948822,
      "rewards/format_reward/mean": 0.12075892686843873,
      "rewards/format_reward/std": 0.3249303221702576,
      "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": 0.021428571428571443,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1998.8,
      "completions/mean_length": 553.5821655273437,
      "completions/mean_terminated_length": 520.8716064453125,
      "completions/min_length": 4.4,
      "completions/min_terminated_length": 4.4,
      "epoch": 0.31503150315031503,
      "grad_norm": 0.3570958375930786,
      "learning_rate": 6e-07,
      "loss": -0.0081,
      "num_tokens": 15812147.0,
      "reward": 0.4677644371986389,
      "reward_std": 0.6031840801239013,
      "rewards/accuracy_reward/mean": 0.25223416090011597,
      "rewards/accuracy_reward/std": 0.4337872326374054,
      "rewards/cosine_scaled_reward/mean": 0.25069683194160464,
      "rewards/cosine_scaled_reward/std": 0.43269214034080505,
      "rewards/format_reward/mean": 0.1816964328289032,
      "rewards/format_reward/std": 0.38075574636459353,
      "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": 0.01696428571428572,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2001.6,
      "completions/mean_length": 557.9491333007812,
      "completions/mean_terminated_length": 532.2440673828125,
      "completions/min_length": 6.6,
      "completions/min_terminated_length": 6.6,
      "epoch": 0.37803780378037805,
      "grad_norm": 0.4451054632663727,
      "learning_rate": 7.249999999999999e-07,
      "loss": 0.0022,
      "num_tokens": 19200175.0,
      "reward": 0.6430311799049377,
      "reward_std": 0.6464242577552796,
      "rewards/accuracy_reward/mean": 0.3037946462631226,
      "rewards/accuracy_reward/std": 0.4588939487934113,
      "rewards/cosine_scaled_reward/mean": 0.3034730076789856,
      "rewards/cosine_scaled_reward/std": 0.45845308899879456,
      "rewards/format_reward/mean": 0.37499999403953554,
      "rewards/format_reward/std": 0.47941267490386963,
      "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": 0.018080357142857117,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2006.4,
      "completions/mean_length": 611.4062866210937,
      "completions/mean_terminated_length": 584.9557250976562,
      "completions/min_length": 39.0,
      "completions/min_terminated_length": 39.0,
      "epoch": 0.44104410441044106,
      "grad_norm": 0.29937848448753357,
      "learning_rate": 8.499999999999999e-07,
      "loss": 0.0221,
      "num_tokens": 22807531.0,
      "reward": 1.002221155166626,
      "reward_std": 0.6414472222328186,
      "rewards/accuracy_reward/mean": 0.4667410671710968,
      "rewards/accuracy_reward/std": 0.49585614800453187,
      "rewards/cosine_scaled_reward/mean": 0.4662725567817688,
      "rewards/cosine_scaled_reward/std": 0.49544560313224795,
      "rewards/format_reward/mean": 0.604687488079071,
      "rewards/format_reward/std": 0.4856111824512482,
      "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": 0.01540178571428572,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2004.0,
      "completions/mean_length": 666.3100830078125,
      "completions/mean_terminated_length": 644.7192993164062,
      "completions/min_length": 40.2,
      "completions/min_terminated_length": 40.2,
      "epoch": 0.504050405040504,
      "grad_norm": 0.20907825231552124,
      "learning_rate": 9.75e-07,
      "loss": 0.0284,
      "num_tokens": 26643040.0,
      "reward": 1.2907849550247192,
      "reward_std": 0.5618968069553375,
      "rewards/accuracy_reward/mean": 0.5892857074737549,
      "rewards/accuracy_reward/std": 0.49005323052406313,
      "rewards/cosine_scaled_reward/mean": 0.5887125849723815,
      "rewards/cosine_scaled_reward/std": 0.48966326117515563,
      "rewards/format_reward/mean": 0.8142857074737548,
      "rewards/format_reward/std": 0.37804563641548156,
      "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": 0.011383928571428581,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1926.8,
      "completions/mean_length": 675.0848388671875,
      "completions/mean_terminated_length": 659.2405029296875,
      "completions/min_length": 74.4,
      "completions/min_terminated_length": 74.4,
      "epoch": 0.5670567056705671,
      "grad_norm": 0.13796494901180267,
      "learning_rate": 9.99686774088805e-07,
      "loss": 0.0281,
      "num_tokens": 30549836.0,
      "reward": 1.4507519245147704,
      "reward_std": 0.4350401759147644,
      "rewards/accuracy_reward/mean": 0.6517857074737549,
      "rewards/accuracy_reward/std": 0.47552146911621096,
      "rewards/cosine_scaled_reward/mean": 0.6508340954780578,
      "rewards/cosine_scaled_reward/std": 0.47517149448394774,
      "rewards/format_reward/mean": 0.9470982074737548,
      "rewards/format_reward/std": 0.21898569464683532,
      "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": 0.015625,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1992.2,
      "completions/mean_length": 730.3631958007812,
      "completions/mean_terminated_length": 709.4870239257813,
      "completions/min_length": 188.6,
      "completions/min_terminated_length": 188.6,
      "epoch": 0.6300630063006301,
      "grad_norm": 0.1085776761174202,
      "learning_rate": 9.984149663879992e-07,
      "loss": 0.0343,
      "num_tokens": 34629415.0,
      "reward": 1.4995574235916138,
      "reward_std": 0.40187769532203677,
      "rewards/accuracy_reward/mean": 0.6754464268684387,
      "rewards/accuracy_reward/std": 0.4675646245479584,
      "rewards/cosine_scaled_reward/mean": 0.6736683487892151,
      "rewards/cosine_scaled_reward/std": 0.4671317100524902,
      "rewards/format_reward/mean": 0.9745535850524902,
      "rewards/format_reward/std": 0.15644133985042571,
      "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": 0.01026785714285714,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1918.8,
      "completions/mean_length": 718.3185546875,
      "completions/mean_terminated_length": 704.5169189453125,
      "completions/min_length": 187.0,
      "completions/min_terminated_length": 187.0,
      "epoch": 0.693069306930693,
      "grad_norm": 0.1311427652835846,
      "learning_rate": 9.961674879848956e-07,
      "loss": 0.0267,
      "num_tokens": 38633034.0,
      "reward": 1.5256200790405274,
      "reward_std": 0.38292278051376344,
      "rewards/accuracy_reward/mean": 0.6888392925262451,
      "rewards/accuracy_reward/std": 0.46227556467056274,
      "rewards/cosine_scaled_reward/mean": 0.6876238822937012,
      "rewards/cosine_scaled_reward/std": 0.4619812607765198,
      "rewards/format_reward/mean": 0.985937488079071,
      "rewards/format_reward/std": 0.11638441979885102,
      "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": 0.01026785714285714,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1962.6,
      "completions/mean_length": 726.8395385742188,
      "completions/mean_terminated_length": 713.1712280273438,
      "completions/min_length": 185.8,
      "completions/min_terminated_length": 185.8,
      "epoch": 0.7560756075607561,
      "grad_norm": 0.10231342166662216,
      "learning_rate": 9.929487384240101e-07,
      "loss": 0.0236,
      "num_tokens": 42664491.0,
      "reward": 1.5491954326629638,
      "reward_std": 0.3630220055580139,
      "rewards/accuracy_reward/mean": 0.7042410612106323,
      "rewards/accuracy_reward/std": 0.4564294159412384,
      "rewards/cosine_scaled_reward/mean": 0.7012924790382385,
      "rewards/cosine_scaled_reward/std": 0.4559535324573517,
      "rewards/format_reward/mean": 0.9886160731315613,
      "rewards/format_reward/std": 0.10382185429334641,
      "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": 0.01450892857142858,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1976.2,
      "completions/mean_length": 732.7027099609375,
      "completions/mean_terminated_length": 713.31240234375,
      "completions/min_length": 193.4,
      "completions/min_terminated_length": 193.4,
      "epoch": 0.819081908190819,
      "grad_norm": 0.10937383025884628,
      "learning_rate": 9.88765018558761e-07,
      "loss": 0.0175,
      "num_tokens": 46741551.0,
      "reward": 1.5278964996337892,
      "reward_std": 0.35590446591377256,
      "rewards/accuracy_reward/mean": 0.6912946343421936,
      "rewards/accuracy_reward/std": 0.46131653189659116,
      "rewards/cosine_scaled_reward/mean": 0.6897214889526367,
      "rewards/cosine_scaled_reward/std": 0.46095741391181944,
      "rewards/format_reward/mean": 0.9834821462631226,
      "rewards/format_reward/std": 0.12617166191339493,
      "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": 0.00892857142857142,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1979.8,
      "completions/mean_length": 721.2656494140625,
      "completions/mean_terminated_length": 709.3232177734375,
      "completions/min_length": 211.8,
      "completions/min_terminated_length": 211.8,
      "epoch": 0.8820882088208821,
      "grad_norm": 0.10150983184576035,
      "learning_rate": 9.836245182172519e-07,
      "loss": 0.0195,
      "num_tokens": 50729557.0,
      "reward": 1.5760997772216796,
      "reward_std": 0.35591434240341185,
      "rewards/accuracy_reward/mean": 0.7227155566215515,
      "rewards/accuracy_reward/std": 0.4474908530712128,
      "rewards/cosine_scaled_reward/mean": 0.7200031995773315,
      "rewards/cosine_scaled_reward/std": 0.44707669615745543,
      "rewards/format_reward/mean": 0.990625,
      "rewards/format_reward/std": 0.09542139172554016,
      "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": 0.011830357142857139,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2008.8,
      "completions/mean_length": 731.1739135742188,
      "completions/mean_terminated_length": 715.3806396484375,
      "completions/min_length": 151.2,
      "completions/min_terminated_length": 151.2,
      "epoch": 0.9450945094509451,
      "grad_norm": 0.10765000432729721,
      "learning_rate": 9.775373001702997e-07,
      "loss": 0.0213,
      "num_tokens": 54802448.0,
      "reward": 1.5546066045761109,
      "reward_std": 0.32919522523880007,
      "rewards/accuracy_reward/mean": 0.7082589507102967,
      "rewards/accuracy_reward/std": 0.4541184425354004,
      "rewards/cosine_scaled_reward/mean": 0.7047487974166871,
      "rewards/cosine_scaled_reward/std": 0.45348618030548093,
      "rewards/format_reward/mean": 0.9879464268684387,
      "rewards/format_reward/std": 0.10762715935707093,
      "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": 0.013587898751733695,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2006.6,
      "completions/mean_length": 748.0619995117188,
      "completions/mean_terminated_length": 730.115478515625,
      "completions/min_length": 196.2,
      "completions/min_terminated_length": 196.2,
      "epoch": 1.0126012601260126,
      "grad_norm": 0.10716311633586884,
      "learning_rate": 9.705152804330871e-07,
      "loss": 0.015,
      "num_tokens": 58996685.0,
      "reward": 1.5426273107528687,
      "reward_std": 0.36739041805267336,
      "rewards/accuracy_reward/mean": 0.7006696462631226,
      "rewards/accuracy_reward/std": 0.45738084316253663,
      "rewards/cosine_scaled_reward/mean": 0.6977543830871582,
      "rewards/cosine_scaled_reward/std": 0.45684285163879396,
      "rewards/format_reward/mean": 0.9861607193946839,
      "rewards/format_reward/std": 0.11394104063510894,
      "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": 0.011607142857142861,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2002.6,
      "completions/mean_length": 723.7663330078125,
      "completions/mean_terminated_length": 708.2584350585937,
      "completions/min_length": 195.8,
      "completions/min_terminated_length": 195.8,
      "epoch": 1.0756075607560756,
      "grad_norm": 0.11011335998773575,
      "learning_rate": 9.625722049390045e-07,
      "loss": 0.0214,
      "num_tokens": 63042006.0,
      "reward": 1.5647123813629151,
      "reward_std": 0.3331080436706543,
      "rewards/accuracy_reward/mean": 0.7145089268684387,
      "rewards/accuracy_reward/std": 0.4507896304130554,
      "rewards/cosine_scaled_reward/mean": 0.7122372031211853,
      "rewards/cosine_scaled_reward/std": 0.4505117475986481,
      "rewards/format_reward/mean": 0.9881696581840516,
      "rewards/format_reward/std": 0.10608766973018646,
      "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": 0.013839285714285698,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2014.6,
      "completions/mean_length": 756.4529418945312,
      "completions/mean_terminated_length": 738.1903930664063,
      "completions/min_length": 226.4,
      "completions/min_terminated_length": 226.4,
      "epoch": 1.1386138613861387,
      "grad_norm": 0.09067358821630478,
      "learning_rate": 9.537236226313388e-07,
      "loss": 0.0201,
      "num_tokens": 67245803.0,
      "reward": 1.5548530101776123,
      "reward_std": 0.3273438990116119,
      "rewards/accuracy_reward/mean": 0.7091517925262452,
      "rewards/accuracy_reward/std": 0.45312303900718687,
      "rewards/cosine_scaled_reward/mean": 0.7056880831718445,
      "rewards/cosine_scaled_reward/std": 0.45253998041152954,
      "rewards/format_reward/mean": 0.9857142806053162,
      "rewards/format_reward/std": 0.11536658853292465,
      "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": 0.009821428571428559,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2014.4,
      "completions/mean_length": 730.8464721679687,
      "completions/mean_terminated_length": 717.7846435546875,
      "completions/min_length": 215.2,
      "completions/min_terminated_length": 215.2,
      "epoch": 1.2016201620162017,
      "grad_norm": 0.11546584218740463,
      "learning_rate": 9.439868550254868e-07,
      "loss": 0.0172,
      "num_tokens": 71285419.0,
      "reward": 1.5703669548034669,
      "reward_std": 0.34961515069007876,
      "rewards/accuracy_reward/mean": 0.7183035731315612,
      "rewards/accuracy_reward/std": 0.4488598704338074,
      "rewards/cosine_scaled_reward/mean": 0.7148409485816956,
      "rewards/cosine_scaled_reward/std": 0.4484779477119446,
      "rewards/format_reward/mean": 0.9892857074737549,
      "rewards/format_reward/std": 0.0997354432940483,
      "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": 0.007366071428571419,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1957.2,
      "completions/mean_length": 717.281494140625,
      "completions/mean_terminated_length": 707.4199096679688,
      "completions/min_length": 215.6,
      "completions/min_terminated_length": 215.6,
      "epoch": 1.2646264626462647,
      "grad_norm": 0.10343993455171585,
      "learning_rate": 9.333809623012763e-07,
      "loss": 0.0156,
      "num_tokens": 75264080.0,
      "reward": 1.6102480173110962,
      "reward_std": 0.30375452637672423,
      "rewards/accuracy_reward/mean": 0.74375,
      "rewards/accuracy_reward/std": 0.43397698402404783,
      "rewards/cosine_scaled_reward/mean": 0.7412548184394836,
      "rewards/cosine_scaled_reward/std": 0.4337835073471069,
      "rewards/format_reward/mean": 0.9917410731315612,
      "rewards/format_reward/std": 0.09024428129196167,
      "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": 0.009821428571428537,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2011.0,
      "completions/mean_length": 733.489111328125,
      "completions/mean_terminated_length": 720.5335571289063,
      "completions/min_length": 193.0,
      "completions/min_terminated_length": 193.0,
      "epoch": 1.3276327632763276,
      "grad_norm": 0.1060086041688919,
      "learning_rate": 9.219267059917681e-07,
      "loss": 0.0199,
      "num_tokens": 79353743.0,
      "reward": 1.5742143154144288,
      "reward_std": 0.3207855045795441,
      "rewards/accuracy_reward/mean": 0.7203125,
      "rewards/accuracy_reward/std": 0.44847235083580017,
      "rewards/cosine_scaled_reward/mean": 0.7178481221199036,
      "rewards/cosine_scaled_reward/std": 0.44811044335365297,
      "rewards/format_reward/mean": 0.9899553537368775,
      "rewards/format_reward/std": 0.09872472882270814,
      "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": 0.006026785714285721,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1975.8,
      "completions/mean_length": 731.3013671875,
      "completions/mean_terminated_length": 723.3466918945312,
      "completions/min_length": 218.6,
      "completions/min_terminated_length": 218.6,
      "epoch": 1.3906390639063906,
      "grad_norm": 0.12040749192237854,
      "learning_rate": 9.096465083415807e-07,
      "loss": 0.0203,
      "num_tokens": 83426645.0,
      "reward": 1.6153923511505126,
      "reward_std": 0.3091603219509125,
      "rewards/accuracy_reward/mean": 0.7464285731315613,
      "rewards/accuracy_reward/std": 0.43417892456054685,
      "rewards/cosine_scaled_reward/mean": 0.7441774845123291,
      "rewards/cosine_scaled_reward/std": 0.43402957916259766,
      "rewards/format_reward/mean": 0.99375,
      "rewards/format_reward/std": 0.07496543601155281,
      "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": 0.012053571428571441,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1982.8,
      "completions/mean_length": 744.0600830078125,
      "completions/mean_terminated_length": 728.1871215820313,
      "completions/min_length": 205.4,
      "completions/min_terminated_length": 205.4,
      "epoch": 1.4536453645364538,
      "grad_norm": 0.12004546821117401,
      "learning_rate": 8.96564408414293e-07,
      "loss": 0.0196,
      "num_tokens": 87563834.0,
      "reward": 1.5543434858322143,
      "reward_std": 0.31985793113708494,
      "rewards/accuracy_reward/mean": 0.7119892239570618,
      "rewards/accuracy_reward/std": 0.45167325139045716,
      "rewards/cosine_scaled_reward/mean": 0.7037603855133057,
      "rewards/cosine_scaled_reward/std": 0.45347819328308103,
      "rewards/format_reward/mean": 0.987500011920929,
      "rewards/format_reward/std": 0.10970796942710877,
      "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": 0.00714285714285714,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2009.0,
      "completions/mean_length": 722.8154296875,
      "completions/mean_terminated_length": 713.2384399414062,
      "completions/min_length": 201.4,
      "completions/min_terminated_length": 201.4,
      "epoch": 1.5166516651665165,
      "grad_norm": 0.09482675045728683,
      "learning_rate": 8.827060150348489e-07,
      "loss": 0.0093,
      "num_tokens": 91570751.0,
      "reward": 1.604229950904846,
      "reward_std": 0.2770054042339325,
      "rewards/accuracy_reward/mean": 0.7392857074737549,
      "rewards/accuracy_reward/std": 0.438806426525116,
      "rewards/cosine_scaled_reward/mean": 0.7370312094688416,
      "rewards/cosine_scaled_reward/std": 0.43855146169662473,
      "rewards/format_reward/mean": 0.9928571343421936,
      "rewards/format_reward/std": 0.08236530497670173,
      "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": 0.012053571428571441,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2035.0,
      "completions/mean_length": 761.4022583007812,
      "completions/mean_terminated_length": 745.6811889648437,
      "completions/min_length": 196.4,
      "completions/min_terminated_length": 196.4,
      "epoch": 1.5796579657965797,
      "grad_norm": 0.10029105842113495,
      "learning_rate": 8.680984566590792e-07,
      "loss": 0.0132,
      "num_tokens": 95783545.0,
      "reward": 1.5971355199813844,
      "reward_std": 0.28857868909835815,
      "rewards/accuracy_reward/mean": 0.7368303656578064,
      "rewards/accuracy_reward/std": 0.43986259698867797,
      "rewards/cosine_scaled_reward/mean": 0.7335565090179443,
      "rewards/cosine_scaled_reward/std": 0.4395099520683289,
      "rewards/format_reward/mean": 0.9870535731315613,
      "rewards/format_reward/std": 0.11282701939344406,
      "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": 0.010267857142857117,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1965.6,
      "completions/mean_length": 726.2194580078125,
      "completions/mean_terminated_length": 712.602392578125,
      "completions/min_length": 215.4,
      "completions/min_terminated_length": 215.4,
      "epoch": 1.6426642664266426,
      "grad_norm": 0.11565880477428436,
      "learning_rate": 8.527703282684765e-07,
      "loss": 0.0187,
      "num_tokens": 99800320.0,
      "reward": 1.6029990196228028,
      "reward_std": 0.29794790744781496,
      "rewards/accuracy_reward/mean": 0.7397321462631226,
      "rewards/accuracy_reward/std": 0.4377941250801086,
      "rewards/cosine_scaled_reward/mean": 0.7374711394309997,
      "rewards/cosine_scaled_reward/std": 0.43750142455101015,
      "rewards/format_reward/mean": 0.989062488079071,
      "rewards/format_reward/std": 0.10165355652570725,
      "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": 0.006696428571428559,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2003.0,
      "completions/mean_length": 733.4511474609375,
      "completions/mean_terminated_length": 724.589501953125,
      "completions/min_length": 170.4,
      "completions/min_terminated_length": 170.4,
      "epoch": 1.7056705670567056,
      "grad_norm": 0.1487206071615219,
      "learning_rate": 8.36751635394175e-07,
      "loss": 0.0119,
      "num_tokens": 103861941.0,
      "reward": 1.6143394470214845,
      "reward_std": 0.2868204951286316,
      "rewards/accuracy_reward/mean": 0.746875,
      "rewards/accuracy_reward/std": 0.43289305567741393,
      "rewards/cosine_scaled_reward/mean": 0.7434109926223755,
      "rewards/cosine_scaled_reward/std": 0.432673978805542,
      "rewards/format_reward/mean": 0.9915178656578064,
      "rewards/format_reward/std": 0.08637134805321693,
      "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": 0.0049107142857142795,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1965.6,
      "completions/mean_length": 711.4899780273438,
      "completions/mean_terminated_length": 704.9090209960938,
      "completions/min_length": 210.0,
      "completions/min_terminated_length": 210.0,
      "epoch": 1.7686768676867688,
      "grad_norm": 0.11649075895547867,
      "learning_rate": 8.200737353797148e-07,
      "loss": 0.0127,
      "num_tokens": 107860712.0,
      "reward": 1.6414124965667725,
      "reward_std": 0.27274273335933685,
      "rewards/accuracy_reward/mean": 0.7636160612106323,
      "rewards/accuracy_reward/std": 0.42368212938308714,
      "rewards/cosine_scaled_reward/mean": 0.7611732006072998,
      "rewards/cosine_scaled_reward/std": 0.4236971378326416,
      "rewards/format_reward/mean": 0.9944196462631225,
      "rewards/format_reward/std": 0.07310284897685052,
      "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": 0.00758928571428572,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1938.4,
      "completions/mean_length": 727.2859619140625,
      "completions/mean_terminated_length": 717.28701171875,
      "completions/min_length": 227.4,
      "completions/min_terminated_length": 227.4,
      "epoch": 1.8316831683168315,
      "grad_norm": 0.1159307211637497,
      "learning_rate": 8.027692759975693e-07,
      "loss": 0.0154,
      "num_tokens": 111983657.0,
      "reward": 1.5885592937469482,
      "reward_std": 0.29645509719848634,
      "rewards/accuracy_reward/mean": 0.7290178418159485,
      "rewards/accuracy_reward/std": 0.4415256381034851,
      "rewards/cosine_scaled_reward/mean": 0.7266719102859497,
      "rewards/cosine_scaled_reward/std": 0.4412645697593689,
      "rewards/format_reward/mean": 0.9924107193946838,
      "rewards/format_reward/std": 0.0823630191385746,
      "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": 0.00758928571428572,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1896.8,
      "completions/mean_length": 724.03017578125,
      "completions/mean_terminated_length": 713.9320678710938,
      "completions/min_length": 218.2,
      "completions/min_terminated_length": 218.2,
      "epoch": 1.8946894689468947,
      "grad_norm": 0.13099786639213562,
      "learning_rate": 7.848721315395974e-07,
      "loss": 0.0131,
      "num_tokens": 116046080.0,
      "reward": 1.6239006757736205,
      "reward_std": 0.29182094931602476,
      "rewards/accuracy_reward/mean": 0.7533482193946839,
      "rewards/accuracy_reward/std": 0.4308763206005096,
      "rewards/cosine_scaled_reward/mean": 0.7493637323379516,
      "rewards/cosine_scaled_reward/std": 0.43060003519058226,
      "rewards/format_reward/mean": 0.9917410731315612,
      "rewards/format_reward/std": 0.08932423293590545,
      "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": 0.0046875,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1962.8,
      "completions/mean_length": 708.4319702148438,
      "completions/mean_terminated_length": 702.11376953125,
      "completions/min_length": 197.4,
      "completions/min_terminated_length": 197.4,
      "epoch": 1.9576957695769577,
      "grad_norm": 0.1514667421579361,
      "learning_rate": 7.664173365065265e-07,
      "loss": 0.0115,
      "num_tokens": 120009631.0,
      "reward": 1.6182810306549071,
      "reward_std": 0.2975043416023254,
      "rewards/accuracy_reward/mean": 0.7477678537368775,
      "rewards/accuracy_reward/std": 0.4321995139122009,
      "rewards/cosine_scaled_reward/mean": 0.7457136750221253,
      "rewards/cosine_scaled_reward/std": 0.4320578873157501,
      "rewards/format_reward/mean": 0.995312488079071,
      "rewards/format_reward/std": 0.0615858756005764,
      "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": 0.0064255374479889275,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1967.2,
      "completions/mean_length": 701.7818725585937,
      "completions/mean_terminated_length": 693.0792114257813,
      "completions/min_length": 189.6,
      "completions/min_terminated_length": 189.6,
      "epoch": 2.025202520252025,
      "grad_norm": 0.1373867690563202,
      "learning_rate": 7.474410170262718e-07,
      "loss": 0.0105,
      "num_tokens": 123923226.0,
      "reward": 1.6074599981307984,
      "reward_std": 0.28035031259059906,
      "rewards/accuracy_reward/mean": 0.7415178656578064,
      "rewards/accuracy_reward/std": 0.43680434226989745,
      "rewards/cosine_scaled_reward/mean": 0.7381341576576232,
      "rewards/cosine_scaled_reward/std": 0.43655965924263,
      "rewards/format_reward/mean": 0.99375,
      "rewards/format_reward/std": 0.07582019194960594,
      "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": 0.005357142857142861,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1936.6,
      "completions/mean_length": 696.6839599609375,
      "completions/mean_terminated_length": 689.4219970703125,
      "completions/min_length": 212.8,
      "completions/min_terminated_length": 212.8,
      "epoch": 2.0882088208820884,
      "grad_norm": 0.13395772874355316,
      "learning_rate": 7.279803201353449e-07,
      "loss": 0.0106,
      "num_tokens": 127847306.0,
      "reward": 1.6294495105743407,
      "reward_std": 0.2632309556007385,
      "rewards/accuracy_reward/mean": 0.7558035731315613,
      "rewards/accuracy_reward/std": 0.4289921045303345,
      "rewards/cosine_scaled_reward/mean": 0.7535416603088378,
      "rewards/cosine_scaled_reward/std": 0.4288810133934021,
      "rewards/format_reward/mean": 0.993749988079071,
      "rewards/format_reward/std": 0.07680554389953613,
      "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": 0.003571428571428581,
      "completions/max_length": 2026.2,
      "completions/max_terminated_length": 1965.2,
      "completions/mean_length": 725.4308471679688,
      "completions/mean_terminated_length": 720.7519287109375,
      "completions/min_length": 225.8,
      "completions/min_terminated_length": 225.8,
      "epoch": 2.151215121512151,
      "grad_norm": 0.14027050137519836,
      "learning_rate": 7.080733410617851e-07,
      "loss": 0.0119,
      "num_tokens": 131877732.0,
      "reward": 1.6384329319000244,
      "reward_std": 0.2705275148153305,
      "rewards/accuracy_reward/mean": 0.7616071462631225,
      "rewards/accuracy_reward/std": 0.42367841601371764,
      "rewards/cosine_scaled_reward/mean": 0.7578925013542175,
      "rewards/cosine_scaled_reward/std": 0.4236116588115692,
      "rewards/format_reward/mean": 0.9957589268684387,
      "rewards/format_reward/std": 0.057058504223823546,
      "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": 0.007366071428571441,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1993.6,
      "completions/mean_length": 743.7538330078125,
      "completions/mean_terminated_length": 734.13173828125,
      "completions/min_length": 226.4,
      "completions/min_terminated_length": 226.4,
      "epoch": 2.2142214221422143,
      "grad_norm": 0.14795565605163574,
      "learning_rate": 6.87759048651962e-07,
      "loss": 0.0141,
      "num_tokens": 135973037.0,
      "reward": 1.5985393047332763,
      "reward_std": 0.2923765182495117,
      "rewards/accuracy_reward/mean": 0.7366071224212647,
      "rewards/accuracy_reward/std": 0.4388765335083008,
      "rewards/cosine_scaled_reward/mean": 0.7316766262054444,
      "rewards/cosine_scaled_reward/std": 0.43860033750534055,
      "rewards/format_reward/mean": 0.9921875,
      "rewards/format_reward/std": 0.08475830480456352,
      "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": 0.005357142857142861,
      "completions/max_length": 2018.0,
      "completions/max_terminated_length": 1964.4,
      "completions/mean_length": 717.5312866210937,
      "completions/mean_terminated_length": 710.5458862304688,
      "completions/min_length": 207.2,
      "completions/min_terminated_length": 207.2,
      "epoch": 2.2772277227722775,
      "grad_norm": 0.12406405061483383,
      "learning_rate": 6.670772090872278e-07,
      "loss": 0.0113,
      "num_tokens": 139942497.0,
      "reward": 1.6446252107620238,
      "reward_std": 0.26751722395420074,
      "rewards/accuracy_reward/mean": 0.7672940850257873,
      "rewards/accuracy_reward/std": 0.4188421845436096,
      "rewards/cosine_scaled_reward/mean": 0.7645792007446289,
      "rewards/cosine_scaled_reward/std": 0.4186910569667816,
      "rewards/format_reward/mean": 0.9944196343421936,
      "rewards/format_reward/std": 0.06284433305263519,
      "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": 0.003125,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1955.4,
      "completions/mean_length": 712.99423828125,
      "completions/mean_terminated_length": 708.8283447265625,
      "completions/min_length": 204.6,
      "completions/min_terminated_length": 204.6,
      "epoch": 2.34023402340234,
      "grad_norm": 0.1320047676563263,
      "learning_rate": 6.460683080397508e-07,
      "loss": 0.0134,
      "num_tokens": 143991319.0,
      "reward": 1.6394832611083985,
      "reward_std": 0.27438428401947024,
      "rewards/accuracy_reward/mean": 0.7629826188087463,
      "rewards/accuracy_reward/std": 0.4248282194137573,
      "rewards/cosine_scaled_reward/mean": 0.7604665756225586,
      "rewards/cosine_scaled_reward/std": 0.42470039129257203,
      "rewards/format_reward/mean": 0.9966517806053161,
      "rewards/format_reward/std": 0.05458408519625664,
      "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": 0.0049107142857142795,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2016.4,
      "completions/mean_length": 693.4261596679687,
      "completions/mean_terminated_length": 686.7283447265625,
      "completions/min_length": 220.8,
      "completions/min_terminated_length": 220.8,
      "epoch": 2.4032403240324034,
      "grad_norm": 0.14782555401325226,
      "learning_rate": 6.247734714199105e-07,
      "loss": 0.0151,
      "num_tokens": 147901684.0,
      "reward": 1.6312123775482177,
      "reward_std": 0.26806690990924836,
      "rewards/accuracy_reward/mean": 0.7569196462631226,
      "rewards/accuracy_reward/std": 0.428833419084549,
      "rewards/cosine_scaled_reward/mean": 0.7550585746765137,
      "rewards/cosine_scaled_reward/std": 0.42870454788208007,
      "rewards/format_reward/mean": 0.9935267925262451,
      "rewards/format_reward/std": 0.07879414483904838,
      "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": 0.005803571428571441,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1876.6,
      "completions/mean_length": 711.0444580078125,
      "completions/mean_terminated_length": 703.2611206054687,
      "completions/min_length": 180.4,
      "completions/min_terminated_length": 180.4,
      "epoch": 2.466246624662466,
      "grad_norm": 0.162477046251297,
      "learning_rate": 6.032343848703969e-07,
      "loss": 0.0154,
      "num_tokens": 151869315.0,
      "reward": 1.639928650856018,
      "reward_std": 0.273574697971344,
      "rewards/accuracy_reward/mean": 0.7631696343421936,
      "rewards/accuracy_reward/std": 0.42458475232124326,
      "rewards/cosine_scaled_reward/mean": 0.7599910497665405,
      "rewards/cosine_scaled_reward/std": 0.4245521605014801,
      "rewards/format_reward/mean": 0.9935267806053162,
      "rewards/format_reward/std": 0.07808763757348061,
      "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": 0.004241071428571441,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1941.8,
      "completions/mean_length": 718.8129760742188,
      "completions/mean_terminated_length": 713.2514038085938,
      "completions/min_length": 214.0,
      "completions/min_terminated_length": 214.0,
      "epoch": 2.5292529252925293,
      "grad_norm": 0.14557433128356934,
      "learning_rate": 5.814932121646073e-07,
      "loss": 0.0066,
      "num_tokens": 155857253.0,
      "reward": 1.6182387113571166,
      "reward_std": 0.28983315229415896,
      "rewards/accuracy_reward/mean": 0.7482142806053161,
      "rewards/accuracy_reward/std": 0.4310362696647644,
      "rewards/cosine_scaled_reward/mean": 0.7462987899780273,
      "rewards/cosine_scaled_reward/std": 0.43085583448410036,
      "rewards/format_reward/mean": 0.99375,
      "rewards/format_reward/std": 0.07288356423377991,
      "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": 0.005357142857142861,
      "completions/max_length": 2010.8,
      "completions/max_terminated_length": 1930.4,
      "completions/mean_length": 700.2652099609375,
      "completions/mean_terminated_length": 693.0643432617187,
      "completions/min_length": 226.6,
      "completions/min_terminated_length": 226.6,
      "epoch": 2.592259225922592,
      "grad_norm": 0.1486082673072815,
      "learning_rate": 5.595925126690801e-07,
      "loss": 0.013,
      "num_tokens": 159736985.0,
      "reward": 1.6559500694274902,
      "reward_std": 0.2599520206451416,
      "rewards/accuracy_reward/mean": 0.7738839268684388,
      "rewards/accuracy_reward/std": 0.41739049553871155,
      "rewards/cosine_scaled_reward/mean": 0.771275007724762,
      "rewards/cosine_scaled_reward/std": 0.41737368106842043,
      "rewards/format_reward/mean": 0.9928571343421936,
      "rewards/format_reward/std": 0.06702591478824615,
      "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": 0.0055803571428571395,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1987.4,
      "completions/mean_length": 712.8966796875,
      "completions/mean_terminated_length": 705.3931274414062,
      "completions/min_length": 189.0,
      "completions/min_terminated_length": 189.0,
      "epoch": 2.6552655265526552,
      "grad_norm": 0.12872540950775146,
      "learning_rate": 5.375751580315354e-07,
      "loss": 0.0121,
      "num_tokens": 163677482.0,
      "reward": 1.6398494243621826,
      "reward_std": 0.2575774312019348,
      "rewards/accuracy_reward/mean": 0.7622767925262451,
      "rewards/accuracy_reward/std": 0.42372516393661497,
      "rewards/cosine_scaled_reward/mean": 0.7609486818313599,
      "rewards/cosine_scaled_reward/std": 0.4236899495124817,
      "rewards/format_reward/mean": 0.9941964387893677,
      "rewards/format_reward/std": 0.07346804365515709,
      "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": 0.0022321428571428826,
      "completions/max_length": 2038.0,
      "completions/max_terminated_length": 1982.6,
      "completions/mean_length": 703.0440185546875,
      "completions/mean_terminated_length": 700.063623046875,
      "completions/min_length": 237.0,
      "completions/min_terminated_length": 237.0,
      "epoch": 2.7182718271827184,
      "grad_norm": 0.11875882744789124,
      "learning_rate": 5.154842482576126e-07,
      "loss": 0.0083,
      "num_tokens": 167635639.0,
      "reward": 1.6419362783432008,
      "reward_std": 0.2711367040872574,
      "rewards/accuracy_reward/mean": 0.7629464268684387,
      "rewards/accuracy_reward/std": 0.42230920791625975,
      "rewards/cosine_scaled_reward/mean": 0.7606581330299378,
      "rewards/cosine_scaled_reward/std": 0.422201132774353,
      "rewards/format_reward/mean": 0.9973214268684387,
      "rewards/format_reward/std": 0.05046965479850769,
      "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": 0.0040178571428571615,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1953.0,
      "completions/mean_length": 718.486181640625,
      "completions/mean_terminated_length": 713.2156860351563,
      "completions/min_length": 237.6,
      "completions/min_terminated_length": 237.6,
      "epoch": 2.781278127812781,
      "grad_norm": 0.14773806929588318,
      "learning_rate": 4.933630273405846e-07,
      "loss": 0.0133,
      "num_tokens": 171601929.0,
      "reward": 1.6592854976654052,
      "reward_std": 0.2805535584688187,
      "rewards/accuracy_reward/mean": 0.7752232193946839,
      "rewards/accuracy_reward/std": 0.41580758690834047,
      "rewards/cosine_scaled_reward/mean": 0.7723655581474305,
      "rewards/cosine_scaled_reward/std": 0.41581328511238097,
      "rewards/format_reward/mean": 0.9957589387893677,
      "rewards/format_reward/std": 0.06227037310600281,
      "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": 0.0033482142857143016,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1964.8,
      "completions/mean_length": 704.973681640625,
      "completions/mean_terminated_length": 700.4504150390625,
      "completions/min_length": 220.8,
      "completions/min_terminated_length": 220.8,
      "epoch": 2.8442844284428443,
      "grad_norm": 0.15314224362373352,
      "learning_rate": 4.712547986092099e-07,
      "loss": 0.0106,
      "num_tokens": 175538803.0,
      "reward": 1.6607633590698243,
      "reward_std": 0.2584164053201675,
      "rewards/accuracy_reward/mean": 0.7758928656578064,
      "rewards/accuracy_reward/std": 0.41488723158836366,
      "rewards/cosine_scaled_reward/mean": 0.7733123421669006,
      "rewards/cosine_scaled_reward/std": 0.41487019062042235,
      "rewards/format_reward/mean": 0.9964285612106323,
      "rewards/format_reward/std": 0.059030205756425855,
      "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": 0.002678571428571441,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1926.0,
      "completions/mean_length": 719.9161010742188,
      "completions/mean_terminated_length": 716.3463256835937,
      "completions/min_length": 232.4,
      "completions/min_terminated_length": 232.4,
      "epoch": 2.9072907290729075,
      "grad_norm": 0.13197660446166992,
      "learning_rate": 4.492028399594338e-07,
      "loss": 0.0071,
      "num_tokens": 179542971.0,
      "reward": 1.6427105903625487,
      "reward_std": 0.2545609176158905,
      "rewards/accuracy_reward/mean": 0.7636160731315613,
      "rewards/accuracy_reward/std": 0.422965270280838,
      "rewards/cosine_scaled_reward/mean": 0.7610907316207886,
      "rewards/cosine_scaled_reward/std": 0.42288472652435305,
      "rewards/format_reward/mean": 0.9970982074737549,
      "rewards/format_reward/std": 0.05258563756942749,
      "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": 0.005803571428571419,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1992.0,
      "completions/mean_length": 749.613427734375,
      "completions/mean_terminated_length": 742.11640625,
      "completions/min_length": 227.6,
      "completions/min_terminated_length": 227.6,
      "epoch": 2.9702970297029703,
      "grad_norm": 0.1521228849887848,
      "learning_rate": 4.2725031913587426e-07,
      "loss": 0.0102,
      "num_tokens": 183700159.0,
      "reward": 1.6156146049499511,
      "reward_std": 0.27087976038455963,
      "rewards/accuracy_reward/mean": 0.7462053656578064,
      "rewards/accuracy_reward/std": 0.4335868239402771,
      "rewards/cosine_scaled_reward/mean": 0.744845175743103,
      "rewards/cosine_scaled_reward/std": 0.43346109986305237,
      "rewards/format_reward/mean": 0.9939731955528259,
      "rewards/format_reward/std": 0.07259788736701012,
      "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": 0.005552184466019417,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1949.6,
      "completions/mean_length": 764.5406005859375,
      "completions/mean_terminated_length": 757.5264038085937,
      "completions/min_length": 240.6,
      "completions/min_terminated_length": 240.6,
      "epoch": 3.037803780378038,
      "grad_norm": 0.1749519556760788,
      "learning_rate": 4.054402092289354e-07,
      "loss": 0.0079,
      "num_tokens": 187977498.0,
      "reward": 1.561311173439026,
      "reward_std": 0.30535221099853516,
      "rewards/accuracy_reward/mean": 0.7107142925262451,
      "rewards/accuracy_reward/std": 0.45136345624923707,
      "rewards/cosine_scaled_reward/mean": 0.7067740797996521,
      "rewards/cosine_scaled_reward/std": 0.45066094398498535,
      "rewards/format_reward/mean": 0.9944196462631225,
      "rewards/format_reward/std": 0.06912030950188637,
      "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": 0.0049107142857142795,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1971.4,
      "completions/mean_length": 733.2201293945312,
      "completions/mean_terminated_length": 726.7033935546875,
      "completions/min_length": 226.6,
      "completions/min_terminated_length": 226.6,
      "epoch": 3.1008100810081007,
      "grad_norm": 0.15469872951507568,
      "learning_rate": 3.838152045529636e-07,
      "loss": 0.0114,
      "num_tokens": 192025564.0,
      "reward": 1.670283031463623,
      "reward_std": 0.2512336015701294,
      "rewards/accuracy_reward/mean": 0.7828125,
      "rewards/accuracy_reward/std": 0.4110615670681,
      "rewards/cosine_scaled_reward/mean": 0.7798516988754273,
      "rewards/cosine_scaled_reward/std": 0.41104503273963927,
      "rewards/format_reward/mean": 0.9950892806053162,
      "rewards/format_reward/std": 0.06898552328348159,
      "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": 0.004910714285714257,
      "completions/max_length": 2043.8,
      "completions/max_terminated_length": 1944.6,
      "completions/mean_length": 732.2582885742188,
      "completions/mean_terminated_length": 725.8395385742188,
      "completions/min_length": 203.8,
      "completions/min_terminated_length": 203.8,
      "epoch": 3.163816381638164,
      "grad_norm": 0.12465714663267136,
      "learning_rate": 3.624176370701235e-07,
      "loss": 0.0125,
      "num_tokens": 196069785.0,
      "reward": 1.6728124141693115,
      "reward_std": 0.26720473170280457,
      "rewards/accuracy_reward/mean": 0.7848214268684387,
      "rewards/accuracy_reward/std": 0.4107992649078369,
      "rewards/cosine_scaled_reward/mean": 0.7813390612602233,
      "rewards/cosine_scaled_reward/std": 0.41090287566184996,
      "rewards/format_reward/mean": 0.9946428537368774,
      "rewards/format_reward/std": 0.061242764443159105,
      "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": 0.004017857142857117,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1911.2,
      "completions/mean_length": 733.9761596679688,
      "completions/mean_terminated_length": 728.7048950195312,
      "completions/min_length": 227.6,
      "completions/min_terminated_length": 227.6,
      "epoch": 3.2268226822682267,
      "grad_norm": 0.1451456993818283,
      "learning_rate": 3.412893935235936e-07,
      "loss": 0.0097,
      "num_tokens": 200116558.0,
      "reward": 1.6486672163009644,
      "reward_std": 0.2642055004835129,
      "rewards/accuracy_reward/mean": 0.7683035731315613,
      "rewards/accuracy_reward/std": 0.4195379137992859,
      "rewards/cosine_scaled_reward/mean": 0.7640753865242005,
      "rewards/cosine_scaled_reward/std": 0.4194532334804535,
      "rewards/format_reward/mean": 0.9966517925262451,
      "rewards/format_reward/std": 0.05103206038475037,
      "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": 0.0046875,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1961.4,
      "completions/mean_length": 744.1857421875,
      "completions/mean_terminated_length": 738.0572265625,
      "completions/min_length": 229.0,
      "completions/min_terminated_length": 229.0,
      "epoch": 3.28982898289829,
      "grad_norm": 0.1736501306295395,
      "learning_rate": 3.204718334422998e-07,
      "loss": 0.0084,
      "num_tokens": 204218390.0,
      "reward": 1.6481477737426757,
      "reward_std": 0.27542482018470765,
      "rewards/accuracy_reward/mean": 0.7685267806053162,
      "rewards/accuracy_reward/std": 0.4199278116226196,
      "rewards/cosine_scaled_reward/mean": 0.7641526460647583,
      "rewards/cosine_scaled_reward/std": 0.4198519289493561,
      "rewards/format_reward/mean": 0.9950892925262451,
      "rewards/format_reward/std": 0.0682831697165966,
      "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": 0.0031250000000000223,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1957.0,
      "completions/mean_length": 741.0214721679688,
      "completions/mean_terminated_length": 736.9272338867188,
      "completions/min_length": 199.4,
      "completions/min_terminated_length": 199.4,
      "epoch": 3.352835283528353,
      "grad_norm": 0.1577436625957489,
      "learning_rate": 3.0000570817769464e-07,
      "loss": 0.0081,
      "num_tokens": 208315302.0,
      "reward": 1.6395839691162108,
      "reward_std": 0.29457778930664064,
      "rewards/accuracy_reward/mean": 0.7618303418159484,
      "rewards/accuracy_reward/std": 0.4226981282234192,
      "rewards/cosine_scaled_reward/mean": 0.7586321473121643,
      "rewards/cosine_scaled_reward/std": 0.42258375883102417,
      "rewards/format_reward/mean": 0.996874988079071,
      "rewards/format_reward/std": 0.05568308532238007,
      "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": 0.004910714285714301,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1973.0,
      "completions/mean_length": 746.5964599609375,
      "completions/mean_terminated_length": 740.2795288085938,
      "completions/min_length": 228.4,
      "completions/min_terminated_length": 228.4,
      "epoch": 3.4158415841584158,
      "grad_norm": 0.16179685294628143,
      "learning_rate": 2.7993108113107443e-07,
      "loss": 0.0063,
      "num_tokens": 212457542.0,
      "reward": 1.6618568658828736,
      "reward_std": 0.25516128838062285,
      "rewards/accuracy_reward/mean": 0.7776785731315613,
      "rewards/accuracy_reward/std": 0.4147783398628235,
      "rewards/cosine_scaled_reward/mean": 0.7734902620315551,
      "rewards/cosine_scaled_reward/std": 0.41475528478622437,
      "rewards/format_reward/mean": 0.9948660731315613,
      "rewards/format_reward/std": 0.06522349417209625,
      "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": 0.00401785714285714,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1937.4,
      "completions/mean_length": 733.0241333007813,
      "completions/mean_terminated_length": 727.7449096679687,
      "completions/min_length": 238.0,
      "completions/min_terminated_length": 238.0,
      "epoch": 3.478847884788479,
      "grad_norm": 0.13407224416732788,
      "learning_rate": 2.602872493275881e-07,
      "loss": 0.0094,
      "num_tokens": 216523066.0,
      "reward": 1.6756507635116578,
      "reward_std": 0.24836473166942596,
      "rewards/accuracy_reward/mean": 0.7861607074737549,
      "rewards/accuracy_reward/std": 0.40959796905517576,
      "rewards/cosine_scaled_reward/mean": 0.7829978227615356,
      "rewards/cosine_scaled_reward/std": 0.4097108542919159,
      "rewards/format_reward/mean": 0.9959821343421936,
      "rewards/format_reward/std": 0.05930030345916748,
      "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": 0.004687499999999978,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1963.0,
      "completions/mean_length": 758.125048828125,
      "completions/mean_terminated_length": 752.0442993164063,
      "completions/min_length": 224.2,
      "completions/min_terminated_length": 224.2,
      "epoch": 3.5418541854185417,
      "grad_norm": 0.24010425806045532,
      "learning_rate": 2.411126664904665e-07,
      "loss": 0.0122,
      "num_tokens": 220698850.0,
      "reward": 1.6234807968139648,
      "reward_std": 0.2710172593593597,
      "rewards/accuracy_reward/mean": 0.7522321462631225,
      "rewards/accuracy_reward/std": 0.4294972479343414,
      "rewards/cosine_scaled_reward/mean": 0.7489704251289367,
      "rewards/cosine_scaled_reward/std": 0.429376620054245,
      "rewards/format_reward/mean": 0.9935267925262451,
      "rewards/format_reward/std": 0.07893114537000656,
      "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": 0.005580357142857118,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1972.6,
      "completions/mean_length": 725.0837524414062,
      "completions/mean_terminated_length": 717.6320190429688,
      "completions/min_length": 235.8,
      "completions/min_terminated_length": 235.8,
      "epoch": 3.604860486048605,
      "grad_norm": 0.15291793644428253,
      "learning_rate": 2.2244486776605241e-07,
      "loss": 0.0062,
      "num_tokens": 224701689.0,
      "reward": 1.6737475872039795,
      "reward_std": 0.24555476903915405,
      "rewards/accuracy_reward/mean": 0.7852678775787354,
      "rewards/accuracy_reward/std": 0.4096054911613464,
      "rewards/cosine_scaled_reward/mean": 0.7825396418571472,
      "rewards/cosine_scaled_reward/std": 0.4097288966178894,
      "rewards/format_reward/mean": 0.9944196224212647,
      "rewards/format_reward/std": 0.07079662829637527,
      "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": 0.0029017857142856983,
      "completions/max_length": 2020.8,
      "completions/max_terminated_length": 1891.0,
      "completions/mean_length": 721.8882080078125,
      "completions/mean_terminated_length": 718.0403076171875,
      "completions/min_length": 218.8,
      "completions/min_terminated_length": 218.8,
      "epoch": 3.667866786678668,
      "grad_norm": 0.17101366817951202,
      "learning_rate": 2.0432039624699015e-07,
      "loss": 0.0122,
      "num_tokens": 228762668.0,
      "reward": 1.6519631147384644,
      "reward_std": 0.2486158311367035,
      "rewards/accuracy_reward/mean": 0.7694196343421936,
      "rewards/accuracy_reward/std": 0.4205454409122467,
      "rewards/cosine_scaled_reward/mean": 0.7682117819786072,
      "rewards/cosine_scaled_reward/std": 0.4205586850643158,
      "rewards/format_reward/mean": 0.996875,
      "rewards/format_reward/std": 0.048269759863615036,
      "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": 0.004910714285714301,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1985.4,
      "completions/mean_length": 755.2917846679687,
      "completions/mean_terminated_length": 748.9348266601562,
      "completions/min_length": 224.4,
      "completions/min_terminated_length": 224.4,
      "epoch": 3.730873087308731,
      "grad_norm": 0.15902669727802277,
      "learning_rate": 1.867747314374041e-07,
      "loss": 0.0129,
      "num_tokens": 232940423.0,
      "reward": 1.6354996681213378,
      "reward_std": 0.25759266018867494,
      "rewards/accuracy_reward/mean": 0.7604910731315613,
      "rewards/accuracy_reward/std": 0.42592642903327943,
      "rewards/cosine_scaled_reward/mean": 0.7549277067184448,
      "rewards/cosine_scaled_reward/std": 0.425705748796463,
      "rewards/format_reward/mean": 0.9950892806053162,
      "rewards/format_reward/std": 0.06402865275740624,
      "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": 0.004464285714285698,
      "completions/max_length": 2028.8,
      "completions/max_terminated_length": 1990.0,
      "completions/mean_length": 745.7913330078125,
      "completions/mean_terminated_length": 739.9611328125,
      "completions/min_length": 235.2,
      "completions/min_terminated_length": 235.2,
      "epoch": 3.793879387938794,
      "grad_norm": 0.16288451850414276,
      "learning_rate": 1.6984221980010454e-07,
      "loss": 0.0103,
      "num_tokens": 237049072.0,
      "reward": 1.678315019607544,
      "reward_std": 0.25401819348335264,
      "rewards/accuracy_reward/mean": 0.7883928537368774,
      "rewards/accuracy_reward/std": 0.4084781646728516,
      "rewards/cosine_scaled_reward/mean": 0.7847549080848694,
      "rewards/cosine_scaled_reward/std": 0.4085943877696991,
      "rewards/format_reward/mean": 0.9950892925262451,
      "rewards/format_reward/std": 0.0597273975610733,
      "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": 0.004464285714285698,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1976.0,
      "completions/mean_length": 747.590234375,
      "completions/mean_terminated_length": 741.756640625,
      "completions/min_length": 220.8,
      "completions/min_terminated_length": 220.8,
      "epoch": 3.856885688568857,
      "grad_norm": 0.16114351153373718,
      "learning_rate": 1.5355600752177312e-07,
      "loss": 0.0113,
      "num_tokens": 241188012.0,
      "reward": 1.6530835628509521,
      "reward_std": 0.2503146827220917,
      "rewards/accuracy_reward/mean": 0.7714285731315613,
      "rewards/accuracy_reward/std": 0.4196752727031708,
      "rewards/cosine_scaled_reward/mean": 0.7677741289138794,
      "rewards/cosine_scaled_reward/std": 0.4196808636188507,
      "rewards/format_reward/mean": 0.9955357074737549,
      "rewards/format_reward/std": 0.06474298015236854,
      "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": 0.005133928571428581,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1974.8,
      "completions/mean_length": 726.7525024414062,
      "completions/mean_terminated_length": 719.9224853515625,
      "completions/min_length": 192.2,
      "completions/min_terminated_length": 192.2,
      "epoch": 3.91989198919892,
      "grad_norm": 0.1752256602048874,
      "learning_rate": 1.3794797562774596e-07,
      "loss": 0.0065,
      "num_tokens": 245204647.0,
      "reward": 1.633379077911377,
      "reward_std": 0.25873309969902036,
      "rewards/accuracy_reward/mean": 0.7604951024055481,
      "rewards/accuracy_reward/std": 0.4261723875999451,
      "rewards/cosine_scaled_reward/mean": 0.7551171183586121,
      "rewards/cosine_scaled_reward/std": 0.42787304520606995,
      "rewards/format_reward/mean": 0.9946428537368774,
      "rewards/format_reward/std": 0.06888841390609741,
      "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": 0.006026785714285698,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2022.2,
      "completions/mean_length": 715.9703491210937,
      "completions/mean_terminated_length": 707.817626953125,
      "completions/min_length": 225.2,
      "completions/min_terminated_length": 225.2,
      "epoch": 3.9828982898289826,
      "grad_norm": 0.1453341394662857,
      "learning_rate": 1.230486775734097e-07,
      "loss": 0.0124,
      "num_tokens": 249253770.0,
      "reward": 1.6327741384506225,
      "reward_std": 0.24864841401576995,
      "rewards/accuracy_reward/mean": 0.7609555959701538,
      "rewards/accuracy_reward/std": 0.42520409226417544,
      "rewards/cosine_scaled_reward/mean": 0.7526016592979431,
      "rewards/cosine_scaled_reward/std": 0.4279878318309784,
      "rewards/format_reward/mean": 0.99375,
      "rewards/format_reward/std": 0.07615087106823921,
      "step": 315
    },
    {
      "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": 0.003649445214979186,
      "completions/max_length": 2018.2,
      "completions/max_terminated_length": 1983.2,
      "completions/mean_length": 726.5947998046875,
      "completions/mean_terminated_length": 721.7894653320312,
      "completions/min_length": 218.0,
      "completions/min_terminated_length": 218.0,
      "epoch": 4.05040504050405,
      "grad_norm": 0.1543412208557129,
      "learning_rate": 1.0888727943437748e-07,
      "loss": 0.0102,
      "num_tokens": 253270490.0,
      "reward": 1.6437525272369384,
      "reward_std": 0.2744287014007568,
      "rewards/accuracy_reward/mean": 0.7688465356826782,
      "rewards/accuracy_reward/std": 0.4200101852416992,
      "rewards/cosine_scaled_reward/mean": 0.7564781427383422,
      "rewards/cosine_scaled_reward/std": 0.4258088171482086,
      "rewards/format_reward/mean": 0.9962053656578064,
      "rewards/format_reward/std": 0.054011975228786466,
      "step": 320
    },
    {
      "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": 0.004464285714285698,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1982.2,
      "completions/mean_length": 722.0165405273438,
      "completions/mean_terminated_length": 716.0443603515625,
      "completions/min_length": 217.0,
      "completions/min_terminated_length": 217.0,
      "epoch": 4.113411341134113,
      "grad_norm": 0.15785960853099823,
      "learning_rate": 9.549150281252632e-08,
      "loss": 0.0099,
      "num_tokens": 257255212.0,
      "reward": 1.6709603309631347,
      "reward_std": 0.26430340111255646,
      "rewards/accuracy_reward/mean": 0.7837053656578064,
      "rewards/accuracy_reward/std": 0.41143102645874025,
      "rewards/cosine_scaled_reward/mean": 0.7803134202957154,
      "rewards/cosine_scaled_reward/std": 0.41151973605155945,
      "rewards/format_reward/mean": 0.9941964387893677,
      "rewards/format_reward/std": 0.07270083874464035,
      "step": 325
    },
    {
      "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": 0.00714285714285714,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1994.2,
      "completions/mean_length": 737.9998168945312,
      "completions/mean_terminated_length": 728.5752319335937,
      "completions/min_length": 211.4,
      "completions/min_terminated_length": 211.4,
      "epoch": 4.176417641764177,
      "grad_norm": 0.15280981361865997,
      "learning_rate": 8.288757056966017e-08,
      "loss": 0.0079,
      "num_tokens": 261348219.0,
      "reward": 1.6478198051452637,
      "reward_std": 0.25256154835224154,
      "rewards/accuracy_reward/mean": 0.7685267806053162,
      "rewards/accuracy_reward/std": 0.4206309378147125,
      "rewards/cosine_scaled_reward/mean": 0.7661752223968505,
      "rewards/cosine_scaled_reward/std": 0.4206216633319855,
      "rewards/format_reward/mean": 0.9924107193946838,
      "rewards/format_reward/std": 0.08273022547364235,
      "step": 330
    },
    {
      "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": 0.006250000000000022,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1988.0,
      "completions/mean_length": 742.7562866210938,
      "completions/mean_terminated_length": 734.589794921875,
      "completions/min_length": 242.2,
      "completions/min_terminated_length": 242.2,
      "epoch": 4.2394239423942395,
      "grad_norm": 0.1629822999238968,
      "learning_rate": 7.110015549502546e-08,
      "loss": 0.0115,
      "num_tokens": 265491519.0,
      "reward": 1.6362776041030884,
      "reward_std": 0.2605998367071152,
      "rewards/accuracy_reward/mean": 0.7607142925262451,
      "rewards/accuracy_reward/std": 0.4256579399108887,
      "rewards/cosine_scaled_reward/mean": 0.7573764562606812,
      "rewards/cosine_scaled_reward/std": 0.42556257247924806,
      "rewards/format_reward/mean": 0.9937500119209289,
      "rewards/format_reward/std": 0.07603473961353302,
      "step": 335
    },
    {
      "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": 0.0020089285714285587,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1959.6,
      "completions/mean_length": 739.0100830078125,
      "completions/mean_terminated_length": 736.3754638671875,
      "completions/min_length": 215.8,
      "completions/min_terminated_length": 215.8,
      "epoch": 4.302430243024302,
      "grad_norm": 0.1339646428823471,
      "learning_rate": 6.015233200716874e-08,
      "loss": 0.0117,
      "num_tokens": 269585804.0,
      "reward": 1.6636471033096314,
      "reward_std": 0.25507272481918336,
      "rewards/accuracy_reward/mean": 0.7779017925262451,
      "rewards/accuracy_reward/std": 0.4153802514076233,
      "rewards/cosine_scaled_reward/mean": 0.7737227082252502,
      "rewards/cosine_scaled_reward/std": 0.4154580056667328,
      "rewards/format_reward/mean": 0.9977678418159485,
      "rewards/format_reward/std": 0.044395358115434644,
      "step": 340
    },
    {
      "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": 0.004464285714285721,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1967.8,
      "completions/mean_length": 720.30205078125,
      "completions/mean_terminated_length": 714.4131469726562,
      "completions/min_length": 203.8,
      "completions/min_terminated_length": 203.8,
      "epoch": 4.365436543654366,
      "grad_norm": 0.18192198872566223,
      "learning_rate": 5.006553098467764e-08,
      "loss": 0.0152,
      "num_tokens": 273613733.0,
      "reward": 1.6808908224105834,
      "reward_std": 0.24185148477554322,
      "rewards/accuracy_reward/mean": 0.7897321343421936,
      "rewards/accuracy_reward/std": 0.4066182732582092,
      "rewards/cosine_scaled_reward/mean": 0.7867815136909485,
      "rewards/cosine_scaled_reward/std": 0.4068154990673065,
      "rewards/format_reward/mean": 0.9955357074737549,
      "rewards/format_reward/std": 0.06201168373227119,
      "step": 345
    },
    {
      "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": 0.003571428571428581,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1949.0,
      "completions/mean_length": 732.9803833007812,
      "completions/mean_terminated_length": 728.2846801757812,
      "completions/min_length": 239.2,
      "completions/min_terminated_length": 239.2,
      "epoch": 4.428442844284429,
      "grad_norm": 0.16086089611053467,
      "learning_rate": 4.085949781423048e-08,
      "loss": 0.0097,
      "num_tokens": 277698637.0,
      "reward": 1.680655264854431,
      "reward_std": 0.23921382129192353,
      "rewards/accuracy_reward/mean": 0.7897321462631226,
      "rewards/accuracy_reward/std": 0.4064399242401123,
      "rewards/cosine_scaled_reward/mean": 0.7863104462623596,
      "rewards/cosine_scaled_reward/std": 0.4065011143684387,
      "rewards/format_reward/mean": 0.9955357193946839,
      "rewards/format_reward/std": 0.06422311514616012,
      "step": 350
    },
    {
      "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": 0.004464285714285721,
      "completions/max_length": 2033.4,
      "completions/max_terminated_length": 1924.2,
      "completions/mean_length": 729.4888793945313,
      "completions/mean_terminated_length": 723.6002075195313,
      "completions/min_length": 190.2,
      "completions/min_terminated_length": 190.2,
      "epoch": 4.491449144914491,
      "grad_norm": 0.1604759842157364,
      "learning_rate": 3.2552253738074074e-08,
      "loss": 0.011,
      "num_tokens": 281751955.0,
      "reward": 1.64648494720459,
      "reward_std": 0.2558329701423645,
      "rewards/accuracy_reward/mean": 0.7671875,
      "rewards/accuracy_reward/std": 0.42172677516937257,
      "rewards/cosine_scaled_reward/mean": 0.7648447513580322,
      "rewards/cosine_scaled_reward/std": 0.4217764616012573,
      "rewards/format_reward/mean": 0.99375,
      "rewards/format_reward/std": 0.0674091026186943,
      "step": 355
    },
    {
      "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": 0.004017857142857117,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1914.6,
      "completions/mean_length": 726.2551635742187,
      "completions/mean_terminated_length": 720.9600830078125,
      "completions/min_length": 226.0,
      "completions/min_terminated_length": 226.0,
      "epoch": 4.554455445544555,
      "grad_norm": 0.16528210043907166,
      "learning_rate": 2.5160060576596365e-08,
      "loss": 0.0093,
      "num_tokens": 285784210.0,
      "reward": 1.644165849685669,
      "reward_std": 0.2550088167190552,
      "rewards/accuracy_reward/mean": 0.7649553537368774,
      "rewards/accuracy_reward/std": 0.4207766890525818,
      "rewards/cosine_scaled_reward/mean": 0.7624387383460999,
      "rewards/cosine_scaled_reward/std": 0.42074838280677795,
      "rewards/format_reward/mean": 0.9959821343421936,
      "rewards/format_reward/std": 0.06139585971832275,
      "step": 360
    },
    {
      "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": 0.00758928571428572,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 2005.2,
      "completions/mean_length": 734.5837524414062,
      "completions/mean_terminated_length": 724.657373046875,
      "completions/min_length": 197.8,
      "completions/min_terminated_length": 197.8,
      "epoch": 4.617461746174618,
      "grad_norm": 0.15532474219799042,
      "learning_rate": 1.8697388895048515e-08,
      "loss": 0.008,
      "num_tokens": 289916185.0,
      "reward": 1.6304625272750854,
      "reward_std": 0.2597217381000519,
      "rewards/accuracy_reward/mean": 0.7575892806053162,
      "rewards/accuracy_reward/std": 0.42815070748329165,
      "rewards/cosine_scaled_reward/mean": 0.7535588622093201,
      "rewards/cosine_scaled_reward/std": 0.4280454754829407,
      "rewards/format_reward/mean": 0.9921875,
      "rewards/format_reward/std": 0.0802665263414383,
      "step": 365
    },
    {
      "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": 0.0020089285714285808,
      "completions/max_length": 2038.2,
      "completions/max_terminated_length": 1860.0,
      "completions/mean_length": 702.7852905273437,
      "completions/mean_terminated_length": 700.0497802734375,
      "completions/min_length": 239.2,
      "completions/min_terminated_length": 239.2,
      "epoch": 4.68046804680468,
      "grad_norm": 0.1871415078639984,
      "learning_rate": 1.3176889676734692e-08,
      "loss": 0.0091,
      "num_tokens": 293819303.0,
      "reward": 1.6794666051864624,
      "reward_std": 0.26247662901878355,
      "rewards/accuracy_reward/mean": 0.7879464268684387,
      "rewards/accuracy_reward/std": 0.408376944065094,
      "rewards/cosine_scaled_reward/mean": 0.7854955077171326,
      "rewards/cosine_scaled_reward/std": 0.4085055232048035,
      "rewards/format_reward/mean": 0.9975446343421936,
      "rewards/format_reward/std": 0.0441232718527317,
      "step": 370
    },
    {
      "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": 0.005580357142857162,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1976.2,
      "completions/mean_length": 738.7837280273437,
      "completions/mean_terminated_length": 731.4327758789062,
      "completions/min_length": 235.6,
      "completions/min_terminated_length": 235.6,
      "epoch": 4.743474347434743,
      "grad_norm": 0.17294685542583466,
      "learning_rate": 8.60936955811753e-09,
      "loss": 0.0117,
      "num_tokens": 297898502.0,
      "reward": 1.675231122970581,
      "reward_std": 0.25796766579151154,
      "rewards/accuracy_reward/mean": 0.7866071343421936,
      "rewards/accuracy_reward/std": 0.408166778087616,
      "rewards/cosine_scaled_reward/mean": 0.7832746028900146,
      "rewards/cosine_scaled_reward/std": 0.40820558071136476,
      "rewards/format_reward/mean": 0.9939732074737548,
      "rewards/format_reward/std": 0.07510094344615936,
      "step": 375
    },
    {
      "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": 0.004464285714285721,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1943.2,
      "completions/mean_length": 707.327490234375,
      "completions/mean_terminated_length": 701.3208129882812,
      "completions/min_length": 241.4,
      "completions/min_terminated_length": 241.4,
      "epoch": 4.806480648064807,
      "grad_norm": 0.1627655327320099,
      "learning_rate": 5.003769674320368e-09,
      "loss": 0.0098,
      "num_tokens": 301826889.0,
      "reward": 1.6730069637298584,
      "reward_std": 0.25484800338745117,
      "rewards/accuracy_reward/mean": 0.7845982193946839,
      "rewards/accuracy_reward/std": 0.40905932784080506,
      "rewards/cosine_scaled_reward/mean": 0.7815048575401307,
      "rewards/cosine_scaled_reward/std": 0.40920759439468385,
      "rewards/format_reward/mean": 0.995312488079071,
      "rewards/format_reward/std": 0.06412766650319099,
      "step": 380
    },
    {
      "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": 0.006026785714285721,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1974.2,
      "completions/mean_length": 738.5777099609375,
      "completions/mean_terminated_length": 730.669482421875,
      "completions/min_length": 202.4,
      "completions/min_terminated_length": 202.4,
      "epoch": 4.8694869486948695,
      "grad_norm": 0.15702560544013977,
      "learning_rate": 2.3671481564347396e-09,
      "loss": 0.0108,
      "num_tokens": 305901773.0,
      "reward": 1.666380262374878,
      "reward_std": 0.25681224167346955,
      "rewards/accuracy_reward/mean": 0.7810267686843873,
      "rewards/accuracy_reward/std": 0.41189714074134826,
      "rewards/cosine_scaled_reward/mean": 0.7774031162261963,
      "rewards/cosine_scaled_reward/std": 0.4118649184703827,
      "rewards/format_reward/mean": 0.9933035731315613,
      "rewards/format_reward/std": 0.07954220175743103,
      "step": 385
    },
    {
      "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": 0.004241071428571441,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1968.4,
      "completions/mean_length": 745.6741455078125,
      "completions/mean_terminated_length": 740.12138671875,
      "completions/min_length": 237.4,
      "completions/min_terminated_length": 237.4,
      "epoch": 4.932493249324932,
      "grad_norm": 0.17807839810848236,
      "learning_rate": 7.046663148979615e-10,
      "loss": 0.0081,
      "num_tokens": 310014129.0,
      "reward": 1.6689030408859253,
      "reward_std": 0.2475011855363846,
      "rewards/accuracy_reward/mean": 0.7819196343421936,
      "rewards/accuracy_reward/std": 0.41213822960853574,
      "rewards/cosine_scaled_reward/mean": 0.7782077431678772,
      "rewards/cosine_scaled_reward/std": 0.4122960567474365,
      "rewards/format_reward/mean": 0.9957589387893677,
      "rewards/format_reward/std": 0.06400550231337547,
      "step": 390
    },
    {
      "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": 0.0043580963938973484,
      "completions/max_length": 2048.0,
      "completions/max_terminated_length": 1954.6,
      "completions/mean_length": 715.9204833984375,
      "completions/mean_terminated_length": 710.0726928710938,
      "completions/min_length": 241.8,
      "completions/min_terminated_length": 241.8,
      "epoch": 4.995499549954996,
      "grad_norm": 0.17002251744270325,
      "learning_rate": 1.95785359849987e-11,
      "loss": 0.0128,
      "num_tokens": 314023514.0,
      "reward": 1.6619688987731933,
      "reward_std": 0.24724579453468323,
      "rewards/accuracy_reward/mean": 0.7774553656578064,
      "rewards/accuracy_reward/std": 0.4148886203765869,
      "rewards/cosine_scaled_reward/mean": 0.7746072888374329,
      "rewards/cosine_scaled_reward/std": 0.4147594332695007,
      "rewards/format_reward/mean": 0.9944196462631225,
      "rewards/format_reward/std": 0.07256761640310287,
      "step": 395
    },
    {
      "epoch": 4.995499549954996,
      "step": 395,
      "total_flos": 0.0,
      "train_loss": 0.012358659607228598,
      "train_runtime": 80233.6203,
      "train_samples_per_second": 0.554,
      "train_steps_per_second": 0.005
    }
  ],
  "logging_steps": 5,
  "max_steps": 395,
  "num_input_tokens_seen": 314023514,
  "num_train_epochs": 5,
  "save_steps": 15,
  "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": 8,
  "trial_name": null,
  "trial_params": null
}