{
  "best_metric": 0.12918958067893982,
  "best_model_checkpoint": "/mnt/nvme5n1p1/trained_grpo_distill_14b_rl_70_s2/v3-20250329-085905/checkpoint-60",
  "epoch": 5.333333333333333,
  "eval_steps": 12,
  "global_step": 64,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "clip_ratio": 0.0,
      "completion_length": 6548.83203125,
      "epoch": 0.08333333333333333,
      "grad_norm": 0.16348209977149963,
      "kl": 0.0,
      "learning_rate": 1.6666666666666667e-05,
      "loss": 0.1545702964067459,
      "memory(GiB)": 179.28,
      "response_clip_ratio": 0.021484375,
      "reward": 0.4017895758152008,
      "reward_std": 0.1705751232802868,
      "rewards/CosineReward": 0.401789553463459,
      "rewards/RepetitionPenalty": 0.0,
      "step": 1,
      "train_speed(iter/s)": 0.000246
    },
    {
      "clip_ratio": 0.0,
      "epoch": 0.16666666666666666,
      "grad_norm": 0.16384358704090118,
      "kl": 0.0,
      "learning_rate": 3.3333333333333335e-05,
      "loss": 0.1545702964067459,
      "memory(GiB)": 179.28,
      "step": 2,
      "train_speed(iter/s)": 0.000475
    },
    {
      "clip_ratio": 1.915595930768177e-05,
      "completion_length": 5978.07421875,
      "epoch": 0.25,
      "grad_norm": 0.1866307407617569,
      "kl": -4.8041343688964844e-05,
      "learning_rate": 5e-05,
      "loss": 0.17502783238887787,
      "memory(GiB)": 179.38,
      "response_clip_ratio": 0.025390625,
      "reward": 0.448335699737072,
      "reward_std": 0.1687266193330288,
      "rewards/CosineReward": 0.44833918660879135,
      "rewards/RepetitionPenalty": -3.4611266528372653e-06,
      "step": 3,
      "train_speed(iter/s)": 0.000362
    },
    {
      "clip_ratio": 1.5297786603696295e-05,
      "epoch": 0.3333333333333333,
      "grad_norm": 0.2006995677947998,
      "kl": -9.1552734375e-05,
      "learning_rate": 6.666666666666667e-05,
      "loss": 0.17240501940250397,
      "memory(GiB)": 179.38,
      "step": 4,
      "train_speed(iter/s)": 0.000475
    },
    {
      "clip_ratio": 2.016552230088564e-05,
      "completion_length": 6285.865234375,
      "epoch": 0.4166666666666667,
      "grad_norm": 0.20385593175888062,
      "kl": 0.0004992485046386719,
      "learning_rate": 8.333333333333334e-05,
      "loss": 0.14679813385009766,
      "memory(GiB)": 179.38,
      "response_clip_ratio": 0.017578125,
      "reward": 0.42602989077568054,
      "reward_std": 0.1678701862692833,
      "rewards/CosineReward": 0.4260316267609596,
      "rewards/RepetitionPenalty": -1.698611413303297e-06,
      "step": 5,
      "train_speed(iter/s)": 0.0004
    },
    {
      "clip_ratio": 0.0007764646143186837,
      "epoch": 0.5,
      "grad_norm": 0.23919148743152618,
      "kl": 0.007171630859375,
      "learning_rate": 0.0001,
      "loss": 0.131611168384552,
      "memory(GiB)": 179.38,
      "step": 6,
      "train_speed(iter/s)": 0.000475
    },
    {
      "clip_ratio": 1.895943137242284e-05,
      "completion_length": 6621.03125,
      "epoch": 0.5833333333333334,
      "grad_norm": 0.4720582664012909,
      "kl": 0.053955078125,
      "learning_rate": 9.991540791356342e-05,
      "loss": 0.1904054880142212,
      "memory(GiB)": 179.38,
      "response_clip_ratio": 0.044921875,
      "reward": 0.335109181702137,
      "reward_std": 0.22131388261914253,
      "rewards/CosineReward": 0.33511025086045265,
      "rewards/RepetitionPenalty": -1.070246014478471e-06,
      "step": 7,
      "train_speed(iter/s)": 0.000418
    },
    {
      "clip_ratio": 0.225050151348114,
      "epoch": 0.6666666666666666,
      "grad_norm": 0.46980902552604675,
      "kl": 0.29248046875,
      "learning_rate": 9.966191788709716e-05,
      "loss": 0.1821543574333191,
      "memory(GiB)": 179.38,
      "step": 8,
      "train_speed(iter/s)": 0.000474
    },
    {
      "clip_ratio": 2.0333556221885374e-05,
      "completion_length": 8788.978515625,
      "epoch": 0.75,
      "grad_norm": 0.3960864245891571,
      "kl": 0.5654296875,
      "learning_rate": 9.924038765061042e-05,
      "loss": 0.20740336179733276,
      "memory(GiB)": 179.38,
      "response_clip_ratio": 0.228515625,
      "reward": 0.013692136853933334,
      "reward_std": 0.1656954512000084,
      "rewards/CosineReward": 0.013692137086763978,
      "rewards/RepetitionPenalty": 0.0,
      "step": 9,
      "train_speed(iter/s)": 0.000428
    },
    {
      "clip_ratio": 0.1824856847524643,
      "epoch": 0.8333333333333334,
      "grad_norm": 0.9183385968208313,
      "kl": 1.3125,
      "learning_rate": 9.865224352899119e-05,
      "loss": 0.2214578539133072,
      "memory(GiB)": 179.38,
      "step": 10,
      "train_speed(iter/s)": 0.000472
    },
    {
      "clip_ratio": 2.737712384259794e-05,
      "completion_length": 9197.515625,
      "epoch": 1.0833333333333333,
      "grad_norm": 0.2475370317697525,
      "kl": 1.103515625,
      "learning_rate": 9.789947561577445e-05,
      "loss": 0.23000311851501465,
      "memory(GiB)": 179.38,
      "response_clip_ratio": 0.234375,
      "reward": -0.06951781921088696,
      "reward_std": 0.12206118181347847,
      "rewards/CosineReward": -0.0695177037268877,
      "rewards/RepetitionPenalty": -1.186342473147306e-07,
      "step": 11,
      "train_speed(iter/s)": 0.000434
    },
    {
      "epoch": 1.1666666666666667,
      "grad_norm": 0.1477218121290207,
      "learning_rate": 9.698463103929542e-05,
      "loss": 0.22107946872711182,
      "memory(GiB)": 179.38,
      "step": 12,
      "train_speed(iter/s)": 0.000471
    },
    {
      "epoch": 1.1666666666666667,
      "eval_clip_ratio": 6.940209914318984e-06,
      "eval_completion_length": 7809.857421875,
      "eval_kl": 1.671875,
      "eval_loss": 0.3713299632072449,
      "eval_response_clip_ratio": 0.1428571492433548,
      "eval_reward": -0.0776652991771698,
      "eval_reward_std": 0.145624041557312,
      "eval_rewards/CosineReward": -0.0776652991771698,
      "eval_rewards/RepetitionPenalty": 0.0,
      "eval_runtime": 1004.5261,
      "eval_samples_per_second": 0.007,
      "eval_steps_per_second": 0.001,
      "step": 12
    },
    {
      "clip_ratio": 0.00031743975137032976,
      "completion_length": 9958.951171875,
      "epoch": 1.25,
      "grad_norm": 0.07723930478096008,
      "kl": 1.4765625,
      "learning_rate": 9.591080534401371e-05,
      "loss": 0.2206912338733673,
      "memory(GiB)": 179.38,
      "response_clip_ratio": 0.306640625,
      "reward": -0.12240881286561489,
      "reward_std": 0.09696152806282043,
      "rewards/CosineReward": -0.12240881659090519,
      "rewards/RepetitionPenalty": 0.0,
      "step": 13,
      "train_speed(iter/s)": 0.000424
    },
    {
      "clip_ratio": 0.021007591392844915,
      "epoch": 1.3333333333333333,
      "grad_norm": 0.43600356578826904,
      "kl": 2.18359375,
      "learning_rate": 9.468163201617062e-05,
      "loss": 0.22104433178901672,
      "memory(GiB)": 179.38,
      "step": 14,
      "train_speed(iter/s)": 0.000454
    },
    {
      "clip_ratio": 2.5281403395638335e-05,
      "completion_length": 9398.857421875,
      "epoch": 1.4166666666666667,
      "grad_norm": 0.07701874524354935,
      "kl": 2.037109375,
      "learning_rate": 9.330127018922194e-05,
      "loss": 0.2514866590499878,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.208984375,
      "reward": -0.10524190217256546,
      "reward_std": 0.10260850936174393,
      "rewards/CosineReward": -0.10524190030992031,
      "rewards/RepetitionPenalty": 0.0,
      "step": 15,
      "train_speed(iter/s)": 0.000429
    },
    {
      "clip_ratio": 0.00025034481586772017,
      "epoch": 1.5,
      "grad_norm": 0.10210710018873215,
      "kl": 2.09765625,
      "learning_rate": 9.177439057064683e-05,
      "loss": 0.24714505672454834,
      "memory(GiB)": 179.4,
      "step": 16,
      "train_speed(iter/s)": 0.000456
    },
    {
      "clip_ratio": 2.9667284252354875e-05,
      "completion_length": 8656.3203125,
      "epoch": 1.5833333333333335,
      "grad_norm": 0.08796831965446472,
      "kl": 1.98828125,
      "learning_rate": 9.01061596377522e-05,
      "loss": 0.25992023944854736,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.1328125,
      "reward": -0.03171869460493326,
      "reward_std": 0.12617995031177998,
      "rewards/CosineReward": -0.03171719750389457,
      "rewards/RepetitionPenalty": -1.4950109061828698e-06,
      "step": 17,
      "train_speed(iter/s)": 0.000434
    },
    {
      "clip_ratio": 0.0004736699193017557,
      "epoch": 1.6666666666666665,
      "grad_norm": 0.1108626052737236,
      "kl": 2.15625,
      "learning_rate": 8.83022221559489e-05,
      "loss": 0.253756582736969,
      "memory(GiB)": 179.4,
      "step": 18,
      "train_speed(iter/s)": 0.000457
    },
    {
      "clip_ratio": 4.969747396899038e-05,
      "completion_length": 7443.119140625,
      "epoch": 1.75,
      "grad_norm": 0.30806148052215576,
      "kl": 1.869140625,
      "learning_rate": 8.636868207865244e-05,
      "loss": 0.314345121383667,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.08203125,
      "reward": 0.0504761952906847,
      "reward_std": 0.17868713662028313,
      "rewards/CosineReward": 0.0504761952906847,
      "rewards/RepetitionPenalty": 0.0,
      "step": 19,
      "train_speed(iter/s)": 0.000437
    },
    {
      "clip_ratio": 0.015013973927125335,
      "epoch": 1.8333333333333335,
      "grad_norm": 0.3818390369415283,
      "kl": 2.44140625,
      "learning_rate": 8.43120818934367e-05,
      "loss": 0.3075486719608307,
      "memory(GiB)": 179.4,
      "step": 20,
      "train_speed(iter/s)": 0.000459
    },
    {
      "clip_ratio": 2.5532435302011436e-05,
      "completion_length": 7361.61328125,
      "epoch": 2.0833333333333335,
      "grad_norm": 0.13797414302825928,
      "kl": 2.1328125,
      "learning_rate": 8.213938048432697e-05,
      "loss": 0.30982470512390137,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.06640625,
      "reward": 0.012603676528669894,
      "reward_std": 0.17831061780452728,
      "rewards/CosineReward": 0.012603679089806974,
      "rewards/RepetitionPenalty": 0.0,
      "step": 21,
      "train_speed(iter/s)": 0.00044
    },
    {
      "clip_ratio": 0.0004507543271756731,
      "epoch": 2.1666666666666665,
      "grad_norm": 0.09445751458406448,
      "kl": 2.38671875,
      "learning_rate": 7.985792958513931e-05,
      "loss": 0.3029288649559021,
      "memory(GiB)": 179.4,
      "step": 22,
      "train_speed(iter/s)": 0.000459
    },
    {
      "clip_ratio": 3.390027268324047e-05,
      "completion_length": 7274.9921875,
      "epoch": 2.25,
      "grad_norm": 0.08438517153263092,
      "kl": 2.24609375,
      "learning_rate": 7.74754489035403e-05,
      "loss": 0.3122837543487549,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.107421875,
      "reward": 0.04000083077698946,
      "reward_std": 0.15720339491963387,
      "rewards/CosineReward": 0.04000122472643852,
      "rewards/RepetitionPenalty": -3.942521118460718e-07,
      "step": 23,
      "train_speed(iter/s)": 0.000442
    },
    {
      "epoch": 2.3333333333333335,
      "grad_norm": 0.22712770104408264,
      "learning_rate": 7.500000000000001e-05,
      "loss": 0.309527724981308,
      "memory(GiB)": 179.4,
      "step": 24,
      "train_speed(iter/s)": 0.00046
    },
    {
      "epoch": 2.3333333333333335,
      "eval_clip_ratio": 2.6606481696944684e-05,
      "eval_completion_length": 8159.00048828125,
      "eval_kl": 2.890625,
      "eval_loss": 0.06887147575616837,
      "eval_response_clip_ratio": 0.0,
      "eval_reward": 0.03657619655132294,
      "eval_reward_std": 0.1121237576007843,
      "eval_rewards/CosineReward": 0.036576349288225174,
      "eval_rewards/RepetitionPenalty": -1.5924378260478989e-07,
      "eval_runtime": 1000.6856,
      "eval_samples_per_second": 0.007,
      "eval_steps_per_second": 0.001,
      "step": 24
    },
    {
      "clip_ratio": 0.0004242509678533679,
      "completion_length": 7330.619140625,
      "epoch": 2.4166666666666665,
      "grad_norm": 0.08223269134759903,
      "kl": 2.5,
      "learning_rate": 7.243995901002312e-05,
      "loss": 0.32226651906967163,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.126953125,
      "reward": 0.04721994558349252,
      "reward_std": 0.17519527859985828,
      "rewards/CosineReward": 0.047220606822520494,
      "rewards/RepetitionPenalty": -6.611797402911179e-07,
      "step": 25,
      "train_speed(iter/s)": 0.000436
    },
    {
      "clip_ratio": 0.00010130415648745839,
      "epoch": 2.5,
      "grad_norm": 0.05986466631293297,
      "kl": 2.40234375,
      "learning_rate": 6.980398830195785e-05,
      "loss": 0.3173108994960785,
      "memory(GiB)": 179.4,
      "step": 26,
      "train_speed(iter/s)": 0.000452
    },
    {
      "clip_ratio": 1.904529290186474e-05,
      "completion_length": 7148.123046875,
      "epoch": 2.5833333333333335,
      "grad_norm": 0.08117522299289703,
      "kl": 2.267578125,
      "learning_rate": 6.710100716628344e-05,
      "loss": 0.29803627729415894,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.05859375,
      "reward": 0.08058818010613322,
      "reward_std": 0.2084643319249153,
      "rewards/CosineReward": 0.08058824483305216,
      "rewards/RepetitionPenalty": -6.841068511675985e-08,
      "step": 27,
      "train_speed(iter/s)": 0.000438
    },
    {
      "clip_ratio": 2.8319021566858282e-05,
      "epoch": 2.6666666666666665,
      "grad_norm": 0.06843555718660355,
      "kl": 2.30859375,
      "learning_rate": 6.434016163555452e-05,
      "loss": 0.2943669557571411,
      "memory(GiB)": 179.4,
      "step": 28,
      "train_speed(iter/s)": 0.000453
    },
    {
      "clip_ratio": 2.377905320827267e-05,
      "completion_length": 7043.7734375,
      "epoch": 2.75,
      "grad_norm": 0.11224047094583511,
      "kl": 2.30859375,
      "learning_rate": 6.153079353712201e-05,
      "loss": 0.3068138062953949,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.064453125,
      "reward": 0.10104330768808722,
      "reward_std": 0.1795201115310192,
      "rewards/CosineReward": 0.10104331048205495,
      "rewards/RepetitionPenalty": 0.0,
      "step": 29,
      "train_speed(iter/s)": 0.00044
    },
    {
      "clip_ratio": 2.678473651940294e-05,
      "epoch": 2.8333333333333335,
      "grad_norm": 0.06063569337129593,
      "kl": 2.48046875,
      "learning_rate": 5.868240888334653e-05,
      "loss": 0.3044006824493408,
      "memory(GiB)": 179.4,
      "step": 30,
      "train_speed(iter/s)": 0.000454
    },
    {
      "clip_ratio": 2.3957579742273083e-05,
      "completion_length": 6917.470703125,
      "epoch": 3.0833333333333335,
      "grad_norm": 0.06954994052648544,
      "kl": 2.52734375,
      "learning_rate": 5.5804645706261514e-05,
      "loss": 0.32018184661865234,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.064453125,
      "reward": 0.11851745843887329,
      "reward_std": 0.1883859969675541,
      "rewards/CosineReward": 0.11851745285093784,
      "rewards/RepetitionPenalty": 0.0,
      "step": 31,
      "train_speed(iter/s)": 0.000442
    },
    {
      "clip_ratio": 3.0191856694727903e-05,
      "epoch": 3.1666666666666665,
      "grad_norm": 0.085055872797966,
      "kl": 2.671875,
      "learning_rate": 5.290724144552379e-05,
      "loss": 0.3171422779560089,
      "memory(GiB)": 179.4,
      "step": 32,
      "train_speed(iter/s)": 0.000455
    },
    {
      "clip_ratio": 4.0645980789122405e-05,
      "completion_length": 7217.87109375,
      "epoch": 3.25,
      "grad_norm": 0.2751259207725525,
      "kl": 3.17578125,
      "learning_rate": 5e-05,
      "loss": 0.33012571930885315,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.06640625,
      "reward": 0.07358059100806713,
      "reward_std": 0.19997451454401016,
      "rewards/CosineReward": 0.0735808894969523,
      "rewards/RepetitionPenalty": -2.9900871822974295e-07,
      "step": 33,
      "train_speed(iter/s)": 0.000443
    },
    {
      "clip_ratio": 0.001576803915668279,
      "epoch": 3.3333333333333335,
      "grad_norm": 0.13159900903701782,
      "kl": 2.92578125,
      "learning_rate": 4.709275855447621e-05,
      "loss": 0.32636967301368713,
      "memory(GiB)": 179.4,
      "step": 34,
      "train_speed(iter/s)": 0.000455
    },
    {
      "clip_ratio": 1.7762126390152844e-05,
      "completion_length": 7016.671875,
      "epoch": 3.4166666666666665,
      "grad_norm": 0.12239609658718109,
      "kl": 2.349609375,
      "learning_rate": 4.4195354293738484e-05,
      "loss": 0.30155760049819946,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.044921875,
      "reward": 0.1327731590718031,
      "reward_std": 0.19624915346503258,
      "rewards/CosineReward": 0.13277401961386204,
      "rewards/RepetitionPenalty": -8.578282404414495e-07,
      "step": 35,
      "train_speed(iter/s)": 0.000444
    },
    {
      "epoch": 3.5,
      "grad_norm": 0.17297780513763428,
      "learning_rate": 4.131759111665349e-05,
      "loss": 0.3010379672050476,
      "memory(GiB)": 179.4,
      "step": 36,
      "train_speed(iter/s)": 0.000456
    },
    {
      "epoch": 3.5,
      "eval_clip_ratio": 3.553860733518377e-05,
      "eval_completion_length": 9360.572265625,
      "eval_kl": 3.46875,
      "eval_loss": 0.3076005280017853,
      "eval_response_clip_ratio": 0.1428571492433548,
      "eval_reward": 0.12024763226509094,
      "eval_reward_std": 0.15955841541290283,
      "eval_rewards/CosineReward": 0.12024761736392975,
      "eval_rewards/RepetitionPenalty": 0.0,
      "eval_runtime": 1031.1183,
      "eval_samples_per_second": 0.007,
      "eval_steps_per_second": 0.001,
      "step": 36
    },
    {
      "clip_ratio": 6.486204142674978e-05,
      "completion_length": 7391.548828125,
      "epoch": 3.5833333333333335,
      "grad_norm": 0.0696391686797142,
      "kl": 2.4580078125,
      "learning_rate": 3.846920646287799e-05,
      "loss": 0.2980685234069824,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.056640625,
      "reward": 0.09776598494499922,
      "reward_std": 0.17968057096004486,
      "rewards/CosineReward": 0.09776598028838634,
      "rewards/RepetitionPenalty": 0.0,
      "step": 37,
      "train_speed(iter/s)": 0.00044
    },
    {
      "clip_ratio": 4.5505223170039244e-05,
      "epoch": 3.6666666666666665,
      "grad_norm": 0.075725257396698,
      "kl": 2.65625,
      "learning_rate": 3.5659838364445505e-05,
      "loss": 0.2973749041557312,
      "memory(GiB)": 179.4,
      "step": 38,
      "train_speed(iter/s)": 0.000451
    },
    {
      "clip_ratio": 2.7227523332840065e-05,
      "completion_length": 6970.59765625,
      "epoch": 3.75,
      "grad_norm": 0.13103771209716797,
      "kl": 2.58203125,
      "learning_rate": 3.289899283371657e-05,
      "loss": 0.3383063077926636,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.0546875,
      "reward": 0.1579855252057314,
      "reward_std": 0.2061951532959938,
      "rewards/CosineReward": 0.15798552427440882,
      "rewards/RepetitionPenalty": 0.0,
      "step": 39,
      "train_speed(iter/s)": 0.000441
    },
    {
      "clip_ratio": 1.819473186515097e-05,
      "epoch": 3.8333333333333335,
      "grad_norm": 0.10087469220161438,
      "kl": 2.63671875,
      "learning_rate": 3.019601169804216e-05,
      "loss": 0.33682429790496826,
      "memory(GiB)": 179.4,
      "step": 40,
      "train_speed(iter/s)": 0.000452
    },
    {
      "clip_ratio": 1.6942344245762797e-05,
      "completion_length": 7164.43359375,
      "epoch": 4.083333333333333,
      "grad_norm": 0.05288735032081604,
      "kl": 2.9453125,
      "learning_rate": 2.7560040989976892e-05,
      "loss": 0.34364235401153564,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.07421875,
      "reward": 0.14892909117043018,
      "reward_std": 0.20389077439904213,
      "rewards/CosineReward": 0.14893037267029285,
      "rewards/RepetitionPenalty": -1.2906025403935928e-06,
      "step": 41,
      "train_speed(iter/s)": 0.000442
    },
    {
      "clip_ratio": 1.7799320403355523e-05,
      "epoch": 4.166666666666667,
      "grad_norm": 0.09818096458911896,
      "kl": 3.0234375,
      "learning_rate": 2.500000000000001e-05,
      "loss": 0.3433087170124054,
      "memory(GiB)": 179.4,
      "step": 42,
      "train_speed(iter/s)": 0.000452
    },
    {
      "clip_ratio": 1.8298997474630596e-05,
      "completion_length": 7555.46875,
      "epoch": 4.25,
      "grad_norm": 0.1601419597864151,
      "kl": 3.34765625,
      "learning_rate": 2.25245510964597e-05,
      "loss": 0.3731963634490967,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.091796875,
      "reward": 0.08685891143977642,
      "reward_std": 0.19358530268073082,
      "rewards/CosineReward": 0.08685892261564732,
      "rewards/RepetitionPenalty": 0.0,
      "step": 43,
      "train_speed(iter/s)": 0.000443
    },
    {
      "clip_ratio": 3.215434617231949e-05,
      "epoch": 4.333333333333333,
      "grad_norm": 0.13790880143642426,
      "kl": 3.265625,
      "learning_rate": 2.0142070414860704e-05,
      "loss": 0.37288910150527954,
      "memory(GiB)": 179.4,
      "step": 44,
      "train_speed(iter/s)": 0.000453
    },
    {
      "clip_ratio": 2.2478410755866207e-05,
      "completion_length": 7659.50390625,
      "epoch": 4.416666666666667,
      "grad_norm": 0.0921853631734848,
      "kl": 3.125,
      "learning_rate": 1.7860619515673033e-05,
      "loss": 0.35845914483070374,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.09765625,
      "reward": 0.09167561889626086,
      "reward_std": 0.2029484249651432,
      "rewards/CosineReward": 0.09167627710849047,
      "rewards/RepetitionPenalty": -6.579870728273818e-07,
      "step": 45,
      "train_speed(iter/s)": 0.000444
    },
    {
      "clip_ratio": 5.8838513723458163e-05,
      "epoch": 4.5,
      "grad_norm": 0.029620319604873657,
      "kl": 2.9921875,
      "learning_rate": 1.5687918106563326e-05,
      "loss": 0.35755494236946106,
      "memory(GiB)": 179.4,
      "step": 46,
      "train_speed(iter/s)": 0.000453
    },
    {
      "clip_ratio": 2.0101413156226045e-05,
      "completion_length": 6990.130859375,
      "epoch": 4.583333333333333,
      "grad_norm": 0.15807992219924927,
      "kl": 2.53515625,
      "learning_rate": 1.3631317921347563e-05,
      "loss": 0.33309173583984375,
      "memory(GiB)": 179.4,
      "response_clip_ratio": 0.072265625,
      "reward": 0.184091467410326,
      "reward_std": 0.20932698994874954,
      "rewards/CosineReward": 0.1840914599597454,
      "rewards/RepetitionPenalty": 0.0,
      "step": 47,
      "train_speed(iter/s)": 0.000445
    },
    {
      "epoch": 4.666666666666667,
      "grad_norm": 0.1702960878610611,
      "learning_rate": 1.1697777844051105e-05,
      "loss": 0.33316949009895325,
      "memory(GiB)": 179.4,
      "step": 48,
      "train_speed(iter/s)": 0.000454
    },
    {
      "epoch": 4.666666666666667,
      "eval_clip_ratio": 4.276269828551449e-05,
      "eval_completion_length": 8430.1435546875,
      "eval_kl": 4.125,
      "eval_loss": 0.32876482605934143,
      "eval_response_clip_ratio": 0.2857142984867096,
      "eval_reward": 0.12817195057868958,
      "eval_reward_std": 0.15863054990768433,
      "eval_rewards/CosineReward": 0.12817198038101196,
      "eval_rewards/RepetitionPenalty": 0.0,
      "eval_runtime": 1017.141,
      "eval_samples_per_second": 0.007,
      "eval_steps_per_second": 0.001,
      "step": 48
    },
    {
      "clip_ratio": 1.9754471395572182e-05,
      "completion_length": 7249.76171875,
      "epoch": 4.083333333333333,
      "grad_norm": 0.10699855536222458,
      "kl": 2.52734375,
      "learning_rate": 7.373001848738202e-05,
      "loss": 0.3186631202697754,
      "memory(GiB)": 185.16,
      "response_clip_ratio": 0.0859375,
      "reward": 0.1530955620110035,
      "reward_std": 0.20569872111082077,
      "rewards/CosineReward": 0.15309555269777775,
      "rewards/RepetitionPenalty": 0.0,
      "step": 49,
      "train_speed(iter/s)": 0.011979
    },
    {
      "clip_ratio": 2.3660143597226124e-05,
      "epoch": 4.166666666666667,
      "grad_norm": 0.08193350583314896,
      "kl": 2.57421875,
      "learning_rate": 7.243995901002312e-05,
      "loss": 0.31817346811294556,
      "memory(GiB)": 185.16,
      "step": 50,
      "train_speed(iter/s)": 0.011806
    },
    {
      "clip_ratio": 2.211639503002516e-05,
      "completion_length": 7120.9375,
      "epoch": 4.25,
      "grad_norm": 0.14047524333000183,
      "kl": 3.2421875,
      "learning_rate": 7.113091308703498e-05,
      "loss": 0.3653872609138489,
      "memory(GiB)": 185.16,
      "response_clip_ratio": 0.1171875,
      "reward": 0.19764085859060287,
      "reward_std": 0.21886689960956573,
      "rewards/CosineReward": 0.1976415440440178,
      "rewards/RepetitionPenalty": -6.893382078487775e-07,
      "step": 51,
      "train_speed(iter/s)": 0.006107
    },
    {
      "clip_ratio": 2.842958019755315e-05,
      "epoch": 4.333333333333333,
      "grad_norm": 0.23579397797584534,
      "kl": 3.4140625,
      "learning_rate": 6.980398830195785e-05,
      "loss": 0.3644198775291443,
      "memory(GiB)": 185.16,
      "step": 52,
      "train_speed(iter/s)": 0.006121
    },
    {
      "clip_ratio": 1.6588395283179125e-05,
      "completion_length": 7284.76171875,
      "epoch": 4.416666666666667,
      "grad_norm": 0.04071468114852905,
      "kl": 3.10546875,
      "learning_rate": 6.846030736563422e-05,
      "loss": 0.36724069714546204,
      "memory(GiB)": 184.19,
      "response_clip_ratio": 0.15625,
      "reward": 0.17882150784134865,
      "reward_std": 0.17632417008280754,
      "rewards/CosineReward": 0.17882288992404938,
      "rewards/RepetitionPenalty": -1.3854046585493052e-06,
      "step": 53,
      "train_speed(iter/s)": 0.013032
    },
    {
      "clip_ratio": 0.0018294498731847852,
      "epoch": 4.5,
      "grad_norm": 0.16932378709316254,
      "kl": 2.8046875,
      "learning_rate": 6.710100716628344e-05,
      "loss": 0.3665623068809509,
      "memory(GiB)": 184.19,
      "step": 54,
      "train_speed(iter/s)": 0.012823
    },
    {
      "clip_ratio": 2.780710246952367e-05,
      "completion_length": 7508.099609375,
      "epoch": 4.583333333333333,
      "grad_norm": 0.04224909096956253,
      "kl": 3.03125,
      "learning_rate": 6.572723780758069e-05,
      "loss": 0.3535935878753662,
      "memory(GiB)": 184.19,
      "response_clip_ratio": 0.162109375,
      "reward": 0.17994043603539467,
      "reward_std": 0.185169517993927,
      "rewards/CosineReward": 0.17994044721126556,
      "rewards/RepetitionPenalty": 0.0,
      "step": 55,
      "train_speed(iter/s)": 0.006615
    },
    {
      "clip_ratio": 2.5482849196123425e-05,
      "epoch": 4.666666666666667,
      "grad_norm": 0.03600788488984108,
      "kl": 3.06640625,
      "learning_rate": 6.434016163555452e-05,
      "loss": 0.35127803683280945,
      "memory(GiB)": 184.19,
      "step": 56,
      "train_speed(iter/s)": 0.006618
    },
    {
      "clip_ratio": 2.373553729739797e-05,
      "completion_length": 7532.986328125,
      "epoch": 4.75,
      "grad_norm": 0.04710998013615608,
      "kl": 3.15625,
      "learning_rate": 6.294095225512603e-05,
      "loss": 0.3812779188156128,
      "memory(GiB)": 184.19,
      "response_clip_ratio": 0.150390625,
      "reward": 0.15369250997900963,
      "reward_std": 0.2217605523765087,
      "rewards/CosineReward": 0.15369419567286968,
      "rewards/RepetitionPenalty": -1.679204729043704e-06,
      "step": 57,
      "train_speed(iter/s)": 0.004522
    },
    {
      "clip_ratio": 2.0617421569113503e-05,
      "epoch": 4.833333333333333,
      "grad_norm": 0.030528903007507324,
      "kl": 3.23046875,
      "learning_rate": 6.153079353712201e-05,
      "loss": 0.3794715702533722,
      "memory(GiB)": 184.19,
      "step": 58,
      "train_speed(iter/s)": 0.004546
    },
    {
      "clip_ratio": 2.46459949266864e-05,
      "completion_length": 7458.419921875,
      "epoch": 5.083333333333333,
      "grad_norm": 0.050327010452747345,
      "kl": 3.34765625,
      "learning_rate": 6.0110878616601904e-05,
      "loss": 0.3843703269958496,
      "memory(GiB)": 184.19,
      "response_clip_ratio": 0.181640625,
      "reward": 0.18363566510379314,
      "reward_std": 0.18902317062020302,
      "rewards/CosineReward": 0.18363658897578716,
      "rewards/RepetitionPenalty": -9.101235036723665e-07,
      "step": 59,
      "train_speed(iter/s)": 0.003495
    },
    {
      "epoch": 5.166666666666667,
      "grad_norm": 0.036718469113111496,
      "learning_rate": 5.868240888334653e-05,
      "loss": 0.3826245665550232,
      "memory(GiB)": 184.19,
      "step": 60,
      "train_speed(iter/s)": 0.003524
    },
    {
      "epoch": 5.166666666666667,
      "eval_clip_ratio": 2.5763809389900416e-05,
      "eval_completion_length": 9660.857421875,
      "eval_kl": 4.59375,
      "eval_loss": 0.2822325825691223,
      "eval_response_clip_ratio": 0.2857142984867096,
      "eval_reward": 0.12918958067893982,
      "eval_reward_std": 0.11928828805685043,
      "eval_rewards/CosineReward": 0.12918956577777863,
      "eval_rewards/RepetitionPenalty": 0.0,
      "eval_runtime": 1017.4637,
      "eval_samples_per_second": 0.007,
      "eval_steps_per_second": 0.001,
      "step": 60
    },
    {
      "clip_ratio": 2.5235839984816266e-05,
      "completion_length": 8022.767578125,
      "epoch": 5.083333333333333,
      "grad_norm": 0.13297545909881592,
      "kl": 3.62109375,
      "learning_rate": 5.724659296536233e-05,
      "loss": 0.39252200722694397,
      "memory(GiB)": 149.46,
      "response_clip_ratio": 0.21875,
      "reward": 0.1098299864679575,
      "reward_std": 0.22277764976024628,
      "rewards/CosineReward": 0.10983069613575935,
      "rewards/RepetitionPenalty": -7.085875495249638e-07,
      "step": 61,
      "train_speed(iter/s)": 0.014869
    },
    {
      "clip_ratio": 0.00028464425849961117,
      "epoch": 5.166666666666667,
      "grad_norm": 0.050255097448825836,
      "kl": 3.34375,
      "learning_rate": 5.5804645706261514e-05,
      "loss": 0.39131754636764526,
      "memory(GiB)": 149.46,
      "step": 62,
      "train_speed(iter/s)": 0.0146
    },
    {
      "clip_ratio": 2.5200109121215064e-05,
      "completion_length": 7439.900390625,
      "epoch": 5.25,
      "grad_norm": 0.24896182119846344,
      "kl": 2.83984375,
      "learning_rate": 5.435778713738292e-05,
      "loss": 0.3810032308101654,
      "memory(GiB)": 149.57,
      "response_clip_ratio": 0.177734375,
      "reward": 0.21993902325630188,
      "reward_std": 0.23494866862893105,
      "rewards/CosineReward": 0.2199418805539608,
      "rewards/RepetitionPenalty": -2.850357986972085e-06,
      "step": 63,
      "train_speed(iter/s)": 0.007539
    },
    {
      "clip_ratio": 2.4266442324005766e-05,
      "epoch": 5.333333333333333,
      "grad_norm": 0.08133991807699203,
      "kl": 3.1015625,
      "learning_rate": 5.290724144552379e-05,
      "loss": 0.3777579069137573,
      "memory(GiB)": 149.57,
      "step": 64,
      "train_speed(iter/s)": 0.007528
    }
  ],
  "logging_steps": 1,
  "max_steps": 120,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 4,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}