{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.9734513274336285,
  "eval_steps": 500,
  "global_step": 169,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 186.12500381469727,
      "epoch": 0.017699115044247787,
      "grad_norm": 2.5752858647583037,
      "kl": 0.0,
      "learning_rate": 8.333333333333333e-08,
      "loss": 0.0,
      "reward": 0.48444899916648865,
      "reward_std": 0.46400773525238037,
      "rewards/correct_code_reward_func": 0.3125,
      "rewards/len_reward_func": 0.17194896191358566,
      "step": 1
    },
    {
      "completion_length": 164.375,
      "epoch": 0.035398230088495575,
      "grad_norm": 10.223894549317365,
      "kl": 0.0,
      "learning_rate": 1.6666666666666665e-07,
      "loss": 0.0,
      "reward": 0.47450903058052063,
      "reward_std": 0.40167468786239624,
      "rewards/correct_code_reward_func": 0.2500000149011612,
      "rewards/len_reward_func": 0.22450900077819824,
      "step": 2
    },
    {
      "completion_length": 164.4791717529297,
      "epoch": 0.05309734513274336,
      "grad_norm": 2.2899995314753827,
      "kl": 4.023313522338867e-05,
      "learning_rate": 2.5e-07,
      "loss": 0.0,
      "reward": 0.37769100069999695,
      "reward_std": 0.5044284015893936,
      "rewards/correct_code_reward_func": 0.2083333432674408,
      "rewards/len_reward_func": 0.16935766488313675,
      "step": 3
    },
    {
      "completion_length": 190.12500762939453,
      "epoch": 0.07079646017699115,
      "grad_norm": 6.009103649538728,
      "kl": 8.815526962280273e-05,
      "learning_rate": 3.333333333333333e-07,
      "loss": 0.0,
      "reward": 0.37105801701545715,
      "reward_std": 0.47604119777679443,
      "rewards/correct_code_reward_func": 0.2083333395421505,
      "rewards/len_reward_func": 0.16272468864917755,
      "step": 4
    },
    {
      "completion_length": 126.10416793823242,
      "epoch": 0.08849557522123894,
      "grad_norm": 2.3931977443517964,
      "kl": 3.49879264831543e-05,
      "learning_rate": 4.1666666666666667e-07,
      "loss": 0.0,
      "reward": 0.5486156344413757,
      "reward_std": 0.40741610527038574,
      "rewards/correct_code_reward_func": 0.3958333358168602,
      "rewards/len_reward_func": 0.15278229862451553,
      "step": 5
    },
    {
      "completion_length": 166.45833587646484,
      "epoch": 0.10619469026548672,
      "grad_norm": 1.5486991749328145,
      "kl": 5.221366882324219e-05,
      "learning_rate": 5e-07,
      "loss": 0.0,
      "reward": 0.5515251755714417,
      "reward_std": 0.47483181953430176,
      "rewards/correct_code_reward_func": 0.2916666716337204,
      "rewards/len_reward_func": 0.25985850393772125,
      "step": 6
    },
    {
      "completion_length": 161.58333587646484,
      "epoch": 0.12389380530973451,
      "grad_norm": 5.714527026195051,
      "kl": 5.5789947509765625e-05,
      "learning_rate": 4.999535676028337e-07,
      "loss": 0.0,
      "reward": 0.46977874636650085,
      "reward_std": 0.461082324385643,
      "rewards/correct_code_reward_func": 0.229166679084301,
      "rewards/len_reward_func": 0.24061208218336105,
      "step": 7
    },
    {
      "completion_length": 161.89583587646484,
      "epoch": 0.1415929203539823,
      "grad_norm": 3.6484747523940286,
      "kl": 0.00010442733764648438,
      "learning_rate": 4.998142876590749e-07,
      "loss": 0.0,
      "reward": 0.2943377196788788,
      "reward_std": 0.3771718442440033,
      "rewards/correct_code_reward_func": 0.1666666679084301,
      "rewards/len_reward_func": 0.12767105549573898,
      "step": 8
    },
    {
      "completion_length": 147.7916717529297,
      "epoch": 0.1592920353982301,
      "grad_norm": 14.356223088928283,
      "kl": 0.00017464160919189453,
      "learning_rate": 4.99582211905537e-07,
      "loss": 0.0,
      "reward": 0.6274224072694778,
      "reward_std": 0.43979279696941376,
      "rewards/correct_code_reward_func": 0.4166666865348816,
      "rewards/len_reward_func": 0.21075571328401566,
      "step": 9
    },
    {
      "completion_length": 188.2916717529297,
      "epoch": 0.17699115044247787,
      "grad_norm": 3.0299854001506494,
      "kl": 8.487701416015625e-05,
      "learning_rate": 4.992574265488882e-07,
      "loss": 0.0,
      "reward": 0.502624899148941,
      "reward_std": 0.35237132012844086,
      "rewards/correct_code_reward_func": 0.3125000149011612,
      "rewards/len_reward_func": 0.19012490659952164,
      "step": 10
    },
    {
      "completion_length": 109.1875,
      "epoch": 0.19469026548672566,
      "grad_norm": 1.8891984084011832,
      "kl": 0.00060272216796875,
      "learning_rate": 4.988400522336303e-07,
      "loss": 0.0,
      "reward": 0.6938449144363403,
      "reward_std": 0.5790197551250458,
      "rewards/correct_code_reward_func": 0.5208333432674408,
      "rewards/len_reward_func": 0.1730116307735443,
      "step": 11
    },
    {
      "completion_length": 171.2916717529297,
      "epoch": 0.21238938053097345,
      "grad_norm": 2.7849518039157664,
      "kl": 0.0006923675537109375,
      "learning_rate": 4.983302439972828e-07,
      "loss": 0.0,
      "reward": 0.4239259362220764,
      "reward_std": 0.48799367249011993,
      "rewards/correct_code_reward_func": 0.20833333395421505,
      "rewards/len_reward_func": 0.21559256315231323,
      "step": 12
    },
    {
      "completion_length": 196.2291717529297,
      "epoch": 0.23008849557522124,
      "grad_norm": 1.392005725111761,
      "kl": 0.000652313232421875,
      "learning_rate": 4.977281912127938e-07,
      "loss": 0.0,
      "reward": 0.5957696735858917,
      "reward_std": 0.5921532511711121,
      "rewards/correct_code_reward_func": 0.4166666716337204,
      "rewards/len_reward_func": 0.17910300940275192,
      "step": 13
    },
    {
      "completion_length": 158.64583587646484,
      "epoch": 0.24778761061946902,
      "grad_norm": 1.5701172166666861,
      "kl": 0.000614166259765625,
      "learning_rate": 4.970341175181955e-07,
      "loss": 0.0,
      "reward": 0.46809761226177216,
      "reward_std": 0.4979168623685837,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.19726425409317017,
      "step": 14
    },
    {
      "completion_length": 147.1041717529297,
      "epoch": 0.26548672566371684,
      "grad_norm": 2.032633618646926,
      "kl": 0.0015106201171875,
      "learning_rate": 4.962482807335314e-07,
      "loss": 0.0,
      "reward": 0.5865300893783569,
      "reward_std": 0.5591276288032532,
      "rewards/correct_code_reward_func": 0.4166666716337204,
      "rewards/len_reward_func": 0.16986342146992683,
      "step": 15
    },
    {
      "completion_length": 177.9791717529297,
      "epoch": 0.2831858407079646,
      "grad_norm": 53.72073405864756,
      "kl": 0.018054962158203125,
      "learning_rate": 4.95370972765087e-07,
      "loss": 0.0,
      "reward": 0.8554503321647644,
      "reward_std": 0.48426124453544617,
      "rewards/correct_code_reward_func": 0.5625000298023224,
      "rewards/len_reward_func": 0.2929503321647644,
      "step": 16
    },
    {
      "completion_length": 187.31250762939453,
      "epoch": 0.3008849557522124,
      "grad_norm": 2.4065750408913034,
      "kl": 0.001964569091796875,
      "learning_rate": 4.944025194969586e-07,
      "loss": 0.0,
      "reward": 0.36077703535556793,
      "reward_std": 0.35280461609363556,
      "rewards/correct_code_reward_func": 0.1041666716337204,
      "rewards/len_reward_func": 0.25661035627126694,
      "step": 17
    },
    {
      "completion_length": 186.14583587646484,
      "epoch": 0.3185840707964602,
      "grad_norm": 20.459541697505053,
      "kl": 0.021259307861328125,
      "learning_rate": 4.933432806700003e-07,
      "loss": 0.0,
      "reward": 0.5004578679800034,
      "reward_std": 0.42624031007289886,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.22962453216314316,
      "step": 18
    },
    {
      "completion_length": 184.2916717529297,
      "epoch": 0.336283185840708,
      "grad_norm": 3.4167543405430436,
      "kl": 0.008358001708984375,
      "learning_rate": 4.921936497481956e-07,
      "loss": 0.0,
      "reward": 0.3476429730653763,
      "reward_std": 0.43119728565216064,
      "rewards/correct_code_reward_func": 0.0833333358168602,
      "rewards/len_reward_func": 0.2643096297979355,
      "step": 19
    },
    {
      "completion_length": 179.87500762939453,
      "epoch": 0.35398230088495575,
      "grad_norm": 2.8250290981037574,
      "kl": 0.0052490234375,
      "learning_rate": 4.909540537725006e-07,
      "loss": 0.0,
      "reward": 0.476805180311203,
      "reward_std": 0.36166805028915405,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.2059718370437622,
      "step": 20
    },
    {
      "completion_length": 148.9375,
      "epoch": 0.37168141592920356,
      "grad_norm": 1.9459373869466972,
      "kl": 0.002933502197265625,
      "learning_rate": 4.896249532022171e-07,
      "loss": 0.0,
      "reward": 0.5851579010486603,
      "reward_std": 0.3626042455434799,
      "rewards/correct_code_reward_func": 0.3333333358168602,
      "rewards/len_reward_func": 0.2518245801329613,
      "step": 21
    },
    {
      "completion_length": 213.8541717529297,
      "epoch": 0.3893805309734513,
      "grad_norm": 0.5828214607615373,
      "kl": 0.0029754638671875,
      "learning_rate": 4.882068417439492e-07,
      "loss": 0.0,
      "reward": 0.5553887784481049,
      "reward_std": 0.5428541898727417,
      "rewards/correct_code_reward_func": 0.291666679084301,
      "rewards/len_reward_func": 0.26372211426496506,
      "step": 22
    },
    {
      "completion_length": 113.43750381469727,
      "epoch": 0.40707964601769914,
      "grad_norm": 1.0373945570716487,
      "kl": 0.0034637451171875,
      "learning_rate": 4.867002461682128e-07,
      "loss": 0.0,
      "reward": 0.49457328021526337,
      "reward_std": 0.38711363077163696,
      "rewards/correct_code_reward_func": 0.2291666716337204,
      "rewards/len_reward_func": 0.26540662348270416,
      "step": 23
    },
    {
      "completion_length": 145.5416717529297,
      "epoch": 0.4247787610619469,
      "grad_norm": 0.8868802158083684,
      "kl": 0.00489044189453125,
      "learning_rate": 4.851057261137608e-07,
      "loss": 0.0,
      "reward": 0.7876249551773071,
      "reward_std": 0.4593852013349533,
      "rewards/correct_code_reward_func": 0.5208333432674408,
      "rewards/len_reward_func": 0.26679155230522156,
      "step": 24
    },
    {
      "completion_length": 223.50000762939453,
      "epoch": 0.4424778761061947,
      "grad_norm": 2.7373544353808152,
      "kl": 0.005214691162109375,
      "learning_rate": 4.83423873879701e-07,
      "loss": 0.0,
      "reward": 0.45352864265441895,
      "reward_std": 0.5192392915487289,
      "rewards/correct_code_reward_func": 0.2291666679084301,
      "rewards/len_reward_func": 0.22436197102069855,
      "step": 25
    },
    {
      "completion_length": 172.56250762939453,
      "epoch": 0.46017699115044247,
      "grad_norm": 1.5744378292792975,
      "kl": 0.0035400390625,
      "learning_rate": 4.816553142054805e-07,
      "loss": 0.0,
      "reward": 0.4348207265138626,
      "reward_std": 0.40823744237422943,
      "rewards/correct_code_reward_func": 0.14583333395421505,
      "rewards/len_reward_func": 0.2889874130487442,
      "step": 26
    },
    {
      "completion_length": 163.5416717529297,
      "epoch": 0.4778761061946903,
      "grad_norm": 37.42131429307232,
      "kl": 0.0279998779296875,
      "learning_rate": 4.798007040388211e-07,
      "loss": 0.0,
      "reward": 0.44616006314754486,
      "reward_std": 0.41116151213645935,
      "rewards/correct_code_reward_func": 0.18750000558793545,
      "rewards/len_reward_func": 0.25866006314754486,
      "step": 27
    },
    {
      "completion_length": 147.1458396911621,
      "epoch": 0.49557522123893805,
      "grad_norm": 3.226533563373661,
      "kl": 0.013214111328125,
      "learning_rate": 4.778607322916895e-07,
      "loss": 0.0,
      "reward": 0.5665057748556137,
      "reward_std": 0.3880573436617851,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.29567238688468933,
      "step": 28
    },
    {
      "completion_length": 127.56250381469727,
      "epoch": 0.5132743362831859,
      "grad_norm": 4.362166865135135,
      "kl": 0.01708984375,
      "learning_rate": 4.758361195843951e-07,
      "loss": 0.0,
      "reward": 0.8003825545310974,
      "reward_std": 0.510110080242157,
      "rewards/correct_code_reward_func": 0.4583333432674408,
      "rewards/len_reward_func": 0.34204918146133423,
      "step": 29
    },
    {
      "completion_length": 173.37500381469727,
      "epoch": 0.5309734513274337,
      "grad_norm": 0.8500372624160994,
      "kl": 0.00797271728515625,
      "learning_rate": 4.7372761797790825e-07,
      "loss": 0.0,
      "reward": 0.37487032264471054,
      "reward_std": 0.32572413980960846,
      "rewards/correct_code_reward_func": 0.1458333432674408,
      "rewards/len_reward_func": 0.22903696447610855,
      "step": 30
    },
    {
      "completion_length": 119.14583587646484,
      "epoch": 0.5486725663716814,
      "grad_norm": 2.2352274664870624,
      "kl": 0.014129638671875,
      "learning_rate": 4.7153601069450146e-07,
      "loss": 0.0,
      "reward": 0.5202069282531738,
      "reward_std": 0.5186183750629425,
      "rewards/correct_code_reward_func": 0.291666679084301,
      "rewards/len_reward_func": 0.22854027152061462,
      "step": 31
    },
    {
      "completion_length": 111.31250381469727,
      "epoch": 0.5663716814159292,
      "grad_norm": 1.1283808844332623,
      "kl": 0.015625,
      "learning_rate": 4.6926211182681294e-07,
      "loss": 0.0,
      "reward": 0.4815867692232132,
      "reward_std": 0.294177770614624,
      "rewards/correct_code_reward_func": 0.1458333432674408,
      "rewards/len_reward_func": 0.3357534259557724,
      "step": 32
    },
    {
      "completion_length": 111.87500381469727,
      "epoch": 0.584070796460177,
      "grad_norm": 15.382447167413178,
      "kl": 0.0740966796875,
      "learning_rate": 4.6690676603544553e-07,
      "loss": 0.0001,
      "reward": 0.30563782900571823,
      "reward_std": 0.5501309335231781,
      "rewards/correct_code_reward_func": 0.1458333395421505,
      "rewards/len_reward_func": 0.15980449318885803,
      "step": 33
    },
    {
      "completion_length": 74.31250190734863,
      "epoch": 0.6017699115044248,
      "grad_norm": 6.16835237121908,
      "kl": 0.038818359375,
      "learning_rate": 4.644708482352092e-07,
      "loss": 0.0,
      "reward": 0.7065277397632599,
      "reward_std": 0.4479583501815796,
      "rewards/correct_code_reward_func": 0.3958333358168602,
      "rewards/len_reward_func": 0.3106943815946579,
      "step": 34
    },
    {
      "completion_length": 96.45833587646484,
      "epoch": 0.6194690265486725,
      "grad_norm": 6.369953012669968,
      "kl": 0.035186767578125,
      "learning_rate": 4.6195526327012623e-07,
      "loss": 0.0,
      "reward": 0.3565668910741806,
      "reward_std": 0.39397968351840973,
      "rewards/correct_code_reward_func": 0.08333333395421505,
      "rewards/len_reward_func": 0.273233562707901,
      "step": 35
    },
    {
      "completion_length": 112.14583587646484,
      "epoch": 0.6371681415929203,
      "grad_norm": 3.3487214552142164,
      "kl": 0.0289306640625,
      "learning_rate": 4.5936094557731806e-07,
      "loss": 0.0,
      "reward": 0.7201341390609741,
      "reward_std": 0.3427349328994751,
      "rewards/correct_code_reward_func": 0.3541666865348816,
      "rewards/len_reward_func": 0.3659674823284149,
      "step": 36
    },
    {
      "completion_length": 101.64583587646484,
      "epoch": 0.6548672566371682,
      "grad_norm": 5.120880381398817,
      "kl": 0.05126953125,
      "learning_rate": 4.566888588399006e-07,
      "loss": 0.0001,
      "reward": 0.5204495191574097,
      "reward_std": 0.3199263662099838,
      "rewards/correct_code_reward_func": 0.1250000037252903,
      "rewards/len_reward_func": 0.39544953405857086,
      "step": 37
    },
    {
      "completion_length": 120.5625,
      "epoch": 0.672566371681416,
      "grad_norm": 10.359850576348157,
      "kl": 0.09923553466796875,
      "learning_rate": 4.5393999562901517e-07,
      "loss": 0.0001,
      "reward": 0.6964526474475861,
      "reward_std": 0.5385148227214813,
      "rewards/correct_code_reward_func": 0.4166666865348816,
      "rewards/len_reward_func": 0.27978597581386566,
      "step": 38
    },
    {
      "completion_length": 110.85417175292969,
      "epoch": 0.6902654867256637,
      "grad_norm": 12.273703285519217,
      "kl": 0.0859375,
      "learning_rate": 4.511153770351288e-07,
      "loss": 0.0001,
      "reward": 0.551011860370636,
      "reward_std": 0.36342084407806396,
      "rewards/correct_code_reward_func": 0.1875000111758709,
      "rewards/len_reward_func": 0.3635118901729584,
      "step": 39
    },
    {
      "completion_length": 68.77083587646484,
      "epoch": 0.7079646017699115,
      "grad_norm": 2.2127761337332514,
      "kl": 0.03765869140625,
      "learning_rate": 4.482160522887403e-07,
      "loss": 0.0,
      "reward": 0.7378278374671936,
      "reward_std": 0.39128121733665466,
      "rewards/correct_code_reward_func": 0.4375000149011612,
      "rewards/len_reward_func": 0.3003278076648712,
      "step": 40
    },
    {
      "completion_length": 135.33333587646484,
      "epoch": 0.7256637168141593,
      "grad_norm": 7.600755671861986,
      "kl": 0.1025390625,
      "learning_rate": 4.4524309837063504e-07,
      "loss": 0.0001,
      "reward": 0.6550733149051666,
      "reward_std": 0.5409570336341858,
      "rewards/correct_code_reward_func": 0.3333333432674408,
      "rewards/len_reward_func": 0.32173997163772583,
      "step": 41
    },
    {
      "completion_length": 87.35416793823242,
      "epoch": 0.7433628318584071,
      "grad_norm": 1.6503719126782672,
      "kl": 0.0264892578125,
      "learning_rate": 4.4219761961182965e-07,
      "loss": 0.0,
      "reward": 0.6292450726032257,
      "reward_std": 0.4537566006183624,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.3584117591381073,
      "step": 42
    },
    {
      "completion_length": 65.29166793823242,
      "epoch": 0.7610619469026548,
      "grad_norm": 4.327974078868625,
      "kl": 0.1019287109375,
      "learning_rate": 4.390807472833584e-07,
      "loss": 0.0001,
      "reward": 0.5142460465431213,
      "reward_std": 0.41541849076747894,
      "rewards/correct_code_reward_func": 0.1875000074505806,
      "rewards/len_reward_func": 0.32674603164196014,
      "step": 43
    },
    {
      "completion_length": 88.62500381469727,
      "epoch": 0.7787610619469026,
      "grad_norm": 2.8856078275542814,
      "kl": 0.068115234375,
      "learning_rate": 4.3589363917605234e-07,
      "loss": 0.0001,
      "reward": 0.48976023495197296,
      "reward_std": 0.39692020416259766,
      "rewards/correct_code_reward_func": 0.1875,
      "rewards/len_reward_func": 0.30226024240255356,
      "step": 44
    },
    {
      "completion_length": 78.75000381469727,
      "epoch": 0.7964601769911505,
      "grad_norm": 1.6857493365819893,
      "kl": 0.04052734375,
      "learning_rate": 4.3263747917046697e-07,
      "loss": 0.0,
      "reward": 0.5636806637048721,
      "reward_std": 0.5081076323986053,
      "rewards/correct_code_reward_func": 0.2500000111758709,
      "rewards/len_reward_func": 0.3136806935071945,
      "step": 45
    },
    {
      "completion_length": 78.50000190734863,
      "epoch": 0.8141592920353983,
      "grad_norm": 1.2755984783224172,
      "kl": 0.0218505859375,
      "learning_rate": 4.2931347679711924e-07,
      "loss": 0.0,
      "reward": 0.6265330016613007,
      "reward_std": 0.38107292354106903,
      "rewards/correct_code_reward_func": 0.2916666716337204,
      "rewards/len_reward_func": 0.33486635982990265,
      "step": 46
    },
    {
      "completion_length": 80.29166793823242,
      "epoch": 0.831858407079646,
      "grad_norm": 1.1240225012044924,
      "kl": 0.0335693359375,
      "learning_rate": 4.259228667871962e-07,
      "loss": 0.0,
      "reward": 0.5936128199100494,
      "reward_std": 0.4675884544849396,
      "rewards/correct_code_reward_func": 0.25,
      "rewards/len_reward_func": 0.34361281991004944,
      "step": 47
    },
    {
      "completion_length": 50.0,
      "epoch": 0.8495575221238938,
      "grad_norm": 8.693611957424073,
      "kl": 0.2470703125,
      "learning_rate": 4.224669086139029e-07,
      "loss": 0.0002,
      "reward": 0.49966277182102203,
      "reward_std": 0.3804493248462677,
      "rewards/correct_code_reward_func": 0.08333333395421505,
      "rewards/len_reward_func": 0.41632944345474243,
      "step": 48
    },
    {
      "completion_length": 92.20833587646484,
      "epoch": 0.8672566371681416,
      "grad_norm": 1.2744927278879081,
      "kl": 0.02978515625,
      "learning_rate": 4.1894688602461913e-07,
      "loss": 0.0,
      "reward": 0.7709261626005173,
      "reward_std": 0.3614697605371475,
      "rewards/correct_code_reward_func": 0.3958333544433117,
      "rewards/len_reward_func": 0.3750927746295929,
      "step": 49
    },
    {
      "completion_length": 127.62500381469727,
      "epoch": 0.8849557522123894,
      "grad_norm": 0.566625300559752,
      "kl": 0.016326904296875,
      "learning_rate": 4.153641065640402e-07,
      "loss": 0.0,
      "reward": 0.5525736510753632,
      "reward_std": 0.42078813910484314,
      "rewards/correct_code_reward_func": 0.20833333395421505,
      "rewards/len_reward_func": 0.34424029290676117,
      "step": 50
    },
    {
      "completion_length": 65.375,
      "epoch": 0.9026548672566371,
      "grad_norm": 12.046039464547933,
      "kl": 0.081298828125,
      "learning_rate": 4.11719901088477e-07,
      "loss": 0.0001,
      "reward": 0.6643719673156738,
      "reward_std": 0.4581465870141983,
      "rewards/correct_code_reward_func": 0.3333333432674408,
      "rewards/len_reward_func": 0.3310386538505554,
      "step": 51
    },
    {
      "completion_length": 103.58333587646484,
      "epoch": 0.9203539823008849,
      "grad_norm": 10.540579232965062,
      "kl": 0.0799560546875,
      "learning_rate": 4.080156232714976e-07,
      "loss": 0.0001,
      "reward": 0.6433776319026947,
      "reward_std": 0.3912515640258789,
      "rewards/correct_code_reward_func": 0.29166667722165585,
      "rewards/len_reward_func": 0.3517109453678131,
      "step": 52
    },
    {
      "completion_length": 62.375003814697266,
      "epoch": 0.9380530973451328,
      "grad_norm": 2.819698340947531,
      "kl": 0.115966796875,
      "learning_rate": 4.0425264910109233e-07,
      "loss": 0.0001,
      "reward": 0.7070481181144714,
      "reward_std": 0.44986239075660706,
      "rewards/correct_code_reward_func": 0.3125,
      "rewards/len_reward_func": 0.39454813301563263,
      "step": 53
    },
    {
      "completion_length": 59.0625,
      "epoch": 0.9557522123893806,
      "grad_norm": 3.22095354432732,
      "kl": 0.1131591796875,
      "learning_rate": 4.004323763685511e-07,
      "loss": 0.0001,
      "reward": 0.5839295089244843,
      "reward_std": 0.26084955781698227,
      "rewards/correct_code_reward_func": 0.14583333395421505,
      "rewards/len_reward_func": 0.43809613585472107,
      "step": 54
    },
    {
      "completion_length": 125.58333969116211,
      "epoch": 0.9734513274336283,
      "grad_norm": 8.551645640447115,
      "kl": 0.15509033203125,
      "learning_rate": 3.9655622414924007e-07,
      "loss": 0.0002,
      "reward": 0.528711274266243,
      "reward_std": 0.3454602509737015,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.3620445877313614,
      "step": 55
    },
    {
      "completion_length": 60.000003814697266,
      "epoch": 0.9911504424778761,
      "grad_norm": 38.727062970707685,
      "kl": 0.20166015625,
      "learning_rate": 3.92625632275474e-07,
      "loss": 0.0002,
      "reward": 0.891117125749588,
      "reward_std": 0.36153580248355865,
      "rewards/correct_code_reward_func": 0.5000000149011612,
      "rewards/len_reward_func": 0.39111708104610443,
      "step": 56
    },
    {
      "completion_length": 91.54167175292969,
      "epoch": 1.0,
      "grad_norm": 38.727062970707685,
      "kl": 0.1865234375,
      "learning_rate": 3.886420608016766e-07,
      "loss": 0.0001,
      "reward": 0.5094150304794312,
      "reward_std": 0.4194638133049011,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.3427482843399048,
      "step": 57
    },
    {
      "completion_length": 96.41666793823242,
      "epoch": 1.0176991150442478,
      "grad_norm": 13.425409986987928,
      "kl": 0.04998779296875,
      "learning_rate": 3.846069894620305e-07,
      "loss": 0.0001,
      "reward": 0.5195242762565613,
      "reward_std": 0.3198210597038269,
      "rewards/correct_code_reward_func": 0.125,
      "rewards/len_reward_func": 0.3945242464542389,
      "step": 58
    },
    {
      "completion_length": 79.3125,
      "epoch": 1.0353982300884956,
      "grad_norm": 1.7429704138905477,
      "kl": 0.0509033203125,
      "learning_rate": 3.8052191712081595e-07,
      "loss": 0.0001,
      "reward": 0.655539482831955,
      "reward_std": 0.2987876646220684,
      "rewards/correct_code_reward_func": 0.1875,
      "rewards/len_reward_func": 0.46803948283195496,
      "step": 59
    },
    {
      "completion_length": 61.187503814697266,
      "epoch": 1.0530973451327434,
      "grad_norm": 7.538322601947877,
      "kl": 0.1343994140625,
      "learning_rate": 3.763883612156441e-07,
      "loss": 0.0001,
      "reward": 0.7049243450164795,
      "reward_std": 0.5336888134479523,
      "rewards/correct_code_reward_func": 0.3125000149011612,
      "rewards/len_reward_func": 0.3924243599176407,
      "step": 60
    },
    {
      "completion_length": 71.45833587646484,
      "epoch": 1.0707964601769913,
      "grad_norm": 1.628031440232102,
      "kl": 0.053955078125,
      "learning_rate": 3.7220785719379095e-07,
      "loss": 0.0001,
      "reward": 0.5856566280126572,
      "reward_std": 0.31656239926815033,
      "rewards/correct_code_reward_func": 0.2083333395421505,
      "rewards/len_reward_func": 0.37732329964637756,
      "step": 61
    },
    {
      "completion_length": 80.64583587646484,
      "epoch": 1.0884955752212389,
      "grad_norm": 0.8807120675882707,
      "kl": 0.05035400390625,
      "learning_rate": 3.6798195794184134e-07,
      "loss": 0.0001,
      "reward": 0.5311767309904099,
      "reward_std": 0.2926383763551712,
      "rewards/correct_code_reward_func": 0.1041666716337204,
      "rewards/len_reward_func": 0.42701007425785065,
      "step": 62
    },
    {
      "completion_length": 112.29167175292969,
      "epoch": 1.1061946902654867,
      "grad_norm": 1.0053639016562932,
      "kl": 0.07916259765625,
      "learning_rate": 3.6371223320885484e-07,
      "loss": 0.0001,
      "reward": 0.4587331861257553,
      "reward_std": 0.33583496510982513,
      "rewards/correct_code_reward_func": 0.08333333395421505,
      "rewards/len_reward_func": 0.3753998577594757,
      "step": 63
    },
    {
      "completion_length": 83.20833587646484,
      "epoch": 1.1238938053097345,
      "grad_norm": 1.7713543449970612,
      "kl": 0.08245849609375,
      "learning_rate": 3.594002690232682e-07,
      "loss": 0.0001,
      "reward": 0.6741357147693634,
      "reward_std": 0.43141913414001465,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.4033023715019226,
      "step": 64
    },
    {
      "completion_length": 84.10416984558105,
      "epoch": 1.1415929203539823,
      "grad_norm": 25.050556888869668,
      "kl": 0.12774658203125,
      "learning_rate": 3.5504766710375047e-07,
      "loss": 0.0001,
      "reward": 0.6237881779670715,
      "reward_std": 0.28890369832515717,
      "rewards/correct_code_reward_func": 0.2083333432674408,
      "rewards/len_reward_func": 0.41545483469963074,
      "step": 65
    },
    {
      "completion_length": 82.27083587646484,
      "epoch": 1.1592920353982301,
      "grad_norm": 1.3229069926746502,
      "kl": 0.0810546875,
      "learning_rate": 3.506560442642299e-07,
      "loss": 0.0001,
      "reward": 0.7232638597488403,
      "reward_std": 0.37123511731624603,
      "rewards/correct_code_reward_func": 0.3541666716337204,
      "rewards/len_reward_func": 0.36909720301628113,
      "step": 66
    },
    {
      "completion_length": 58.00000190734863,
      "epoch": 1.176991150442478,
      "grad_norm": 14.439064098343355,
      "kl": 0.1383056640625,
      "learning_rate": 3.462270318133136e-07,
      "loss": 0.0001,
      "reward": 0.9435772597789764,
      "reward_std": 0.2863286882638931,
      "rewards/correct_code_reward_func": 0.520833358168602,
      "rewards/len_reward_func": 0.42274393141269684,
      "step": 67
    },
    {
      "completion_length": 71.10416984558105,
      "epoch": 1.1946902654867257,
      "grad_norm": 7.987124238678158,
      "kl": 0.122314453125,
      "learning_rate": 3.41762274948323e-07,
      "loss": 0.0001,
      "reward": 0.7681821584701538,
      "reward_std": 0.3728492707014084,
      "rewards/correct_code_reward_func": 0.3541666865348816,
      "rewards/len_reward_func": 0.41401544213294983,
      "step": 68
    },
    {
      "completion_length": 60.66666793823242,
      "epoch": 1.2123893805309733,
      "grad_norm": 2.9983488159761817,
      "kl": 0.0537109375,
      "learning_rate": 3.372634321441702e-07,
      "loss": 0.0001,
      "reward": 0.7418330609798431,
      "reward_std": 0.38279734551906586,
      "rewards/correct_code_reward_func": 0.3541666865348816,
      "rewards/len_reward_func": 0.38766637444496155,
      "step": 69
    },
    {
      "completion_length": 80.62500381469727,
      "epoch": 1.2300884955752212,
      "grad_norm": 3.023771746536167,
      "kl": 0.08154296875,
      "learning_rate": 3.3273217453730205e-07,
      "loss": 0.0001,
      "reward": 0.6804832518100739,
      "reward_std": 0.4174923449754715,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.40964992344379425,
      "step": 70
    },
    {
      "completion_length": 69.5,
      "epoch": 1.247787610619469,
      "grad_norm": 1.4050656113052118,
      "kl": 0.0853271484375,
      "learning_rate": 3.281701853049416e-07,
      "loss": 0.0001,
      "reward": 0.7291666865348816,
      "reward_std": 0.32618680596351624,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.4583333432674408,
      "step": 71
    },
    {
      "completion_length": 74.47916793823242,
      "epoch": 1.2654867256637168,
      "grad_norm": 2.8179580865899894,
      "kl": 0.0390625,
      "learning_rate": 3.23579159039856e-07,
      "loss": 0.0,
      "reward": 0.75091952085495,
      "reward_std": 0.4079275578260422,
      "rewards/correct_code_reward_func": 0.2916666716337204,
      "rewards/len_reward_func": 0.45925286412239075,
      "step": 72
    },
    {
      "completion_length": 65.52083587646484,
      "epoch": 1.2831858407079646,
      "grad_norm": 1.1299435463555585,
      "kl": 0.0399169921875,
      "learning_rate": 3.1896080112088477e-07,
      "loss": 0.0,
      "reward": 0.7291666865348816,
      "reward_std": 0.40629878640174866,
      "rewards/correct_code_reward_func": 0.2708333395421505,
      "rewards/len_reward_func": 0.4583333432674408,
      "step": 73
    },
    {
      "completion_length": 57.000003814697266,
      "epoch": 1.3008849557522124,
      "grad_norm": 24.705642459397488,
      "kl": 0.17333984375,
      "learning_rate": 3.1431682707946117e-07,
      "loss": 0.0002,
      "reward": 0.5623736083507538,
      "reward_std": 0.24504420161247253,
      "rewards/correct_code_reward_func": 0.12500000558793545,
      "rewards/len_reward_func": 0.4373735785484314,
      "step": 74
    },
    {
      "completion_length": 78.33333587646484,
      "epoch": 1.3185840707964602,
      "grad_norm": 0.7830442669221305,
      "kl": 0.0224609375,
      "learning_rate": 3.096489619623621e-07,
      "loss": 0.0,
      "reward": 0.7488380074501038,
      "reward_std": 0.30931201577186584,
      "rewards/correct_code_reward_func": 0.2500000149011612,
      "rewards/len_reward_func": 0.49883799254894257,
      "step": 75
    },
    {
      "completion_length": 87.60417175292969,
      "epoch": 1.336283185840708,
      "grad_norm": 2.1746791001290044,
      "kl": 0.04168701171875,
      "learning_rate": 3.049589396909239e-07,
      "loss": 0.0,
      "reward": 0.5558712333440781,
      "reward_std": 0.20058585330843925,
      "rewards/correct_code_reward_func": 0.1041666716337204,
      "rewards/len_reward_func": 0.45170454680919647,
      "step": 76
    },
    {
      "completion_length": 72.14583587646484,
      "epoch": 1.3539823008849559,
      "grad_norm": 0.41651138876768207,
      "kl": 0.0919189453125,
      "learning_rate": 3.0024850241696126e-07,
      "loss": 0.0001,
      "reward": 0.6021915674209595,
      "reward_std": 0.22243696451187134,
      "rewards/correct_code_reward_func": 0.1458333358168602,
      "rewards/len_reward_func": 0.4563582092523575,
      "step": 77
    },
    {
      "completion_length": 54.916666984558105,
      "epoch": 1.3716814159292037,
      "grad_norm": 8.88612982008444,
      "kl": 0.106689453125,
      "learning_rate": 2.9551939987562865e-07,
      "loss": 0.0001,
      "reward": 0.6803071200847626,
      "reward_std": 0.2964301370084286,
      "rewards/correct_code_reward_func": 0.2291666679084301,
      "rewards/len_reward_func": 0.4511404484510422,
      "step": 78
    },
    {
      "completion_length": 93.91666793823242,
      "epoch": 1.3893805309734513,
      "grad_norm": 3.7969039322604865,
      "kl": 0.086761474609375,
      "learning_rate": 2.9077338873546564e-07,
      "loss": 0.0001,
      "reward": 0.7916666865348816,
      "reward_std": 0.34503278136253357,
      "rewards/correct_code_reward_func": 0.2916666716337204,
      "rewards/len_reward_func": 0.5,
      "step": 79
    },
    {
      "completion_length": 56.08333396911621,
      "epoch": 1.407079646017699,
      "grad_norm": 3.0759876398616743,
      "kl": 0.076171875,
      "learning_rate": 2.860122319458661e-07,
      "loss": 0.0001,
      "reward": 0.8750000596046448,
      "reward_std": 0.36751919984817505,
      "rewards/correct_code_reward_func": 0.3958333432674408,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 80
    },
    {
      "completion_length": 77.58333587646484,
      "epoch": 1.424778761061947,
      "grad_norm": 5.267849973823214,
      "kl": 0.0604248046875,
      "learning_rate": 2.8123769808221403e-07,
      "loss": 0.0001,
      "reward": 0.9166666865348816,
      "reward_std": 0.4720035046339035,
      "rewards/correct_code_reward_func": 0.4375000298023224,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 81
    },
    {
      "completion_length": 86.18750190734863,
      "epoch": 1.4424778761061947,
      "grad_norm": 16.256704067320886,
      "kl": 0.05712890625,
      "learning_rate": 2.764515606889307e-07,
      "loss": 0.0001,
      "reward": 0.8077563941478729,
      "reward_std": 0.3568233400583267,
      "rewards/correct_code_reward_func": 0.3958333432674408,
      "rewards/len_reward_func": 0.41192303597927094,
      "step": 82
    },
    {
      "completion_length": 106.89583587646484,
      "epoch": 1.4601769911504425,
      "grad_norm": 2.2341312424056214,
      "kl": 0.021087646484375,
      "learning_rate": 2.716555976206747e-07,
      "loss": 0.0,
      "reward": 0.7582251131534576,
      "reward_std": 0.3638576865196228,
      "rewards/correct_code_reward_func": 0.2916666716337204,
      "rewards/len_reward_func": 0.46655844151973724,
      "step": 83
    },
    {
      "completion_length": 48.41666793823242,
      "epoch": 1.4778761061946903,
      "grad_norm": 1.3490761289725859,
      "kl": 0.05279541015625,
      "learning_rate": 2.66851590381942e-07,
      "loss": 0.0001,
      "reward": 0.7612736523151398,
      "reward_std": 0.23646536469459534,
      "rewards/correct_code_reward_func": 0.2708333358168602,
      "rewards/len_reward_func": 0.490440309047699,
      "step": 84
    },
    {
      "completion_length": 60.187503814697266,
      "epoch": 1.495575221238938,
      "grad_norm": 4.538439276716702,
      "kl": 0.0347900390625,
      "learning_rate": 2.620413234653093e-07,
      "loss": 0.0,
      "reward": 0.8333333432674408,
      "reward_std": 0.438393235206604,
      "rewards/correct_code_reward_func": 0.3541666716337204,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 85
    },
    {
      "completion_length": 138.50000762939453,
      "epoch": 1.5132743362831858,
      "grad_norm": 0.9112971760232713,
      "kl": 0.021392822265625,
      "learning_rate": 2.5722658368856814e-07,
      "loss": 0.0,
      "reward": 0.6596881151199341,
      "reward_std": 0.4098925292491913,
      "rewards/correct_code_reward_func": 0.1875000074505806,
      "rewards/len_reward_func": 0.4721881300210953,
      "step": 86
    },
    {
      "completion_length": 67.89583587646484,
      "epoch": 1.5309734513274336,
      "grad_norm": 1.5750222049346805,
      "kl": 0.033447265625,
      "learning_rate": 2.5240915953099515e-07,
      "loss": 0.0,
      "reward": 0.9335185885429382,
      "reward_std": 0.32536589354276657,
      "rewards/correct_code_reward_func": 0.4375000149011612,
      "rewards/len_reward_func": 0.49601852893829346,
      "step": 87
    },
    {
      "completion_length": 56.395835876464844,
      "epoch": 1.5486725663716814,
      "grad_norm": 0.8224337332016152,
      "kl": 0.0401611328125,
      "learning_rate": 2.4759084046900483e-07,
      "loss": 0.0,
      "reward": 0.7500000298023224,
      "reward_std": 0.2342708334326744,
      "rewards/correct_code_reward_func": 0.3125000149011612,
      "rewards/len_reward_func": 0.4375000149011612,
      "step": 88
    },
    {
      "completion_length": 55.250003814697266,
      "epoch": 1.5663716814159292,
      "grad_norm": 1.7919585934332407,
      "kl": 0.046142578125,
      "learning_rate": 2.427734163114319e-07,
      "loss": 0.0,
      "reward": 0.8125000298023224,
      "reward_std": 0.3177132308483124,
      "rewards/correct_code_reward_func": 0.3125000149011612,
      "rewards/len_reward_func": 0.5,
      "step": 89
    },
    {
      "completion_length": 78.27083587646484,
      "epoch": 1.584070796460177,
      "grad_norm": 1.6894819978109317,
      "kl": 0.0758056640625,
      "learning_rate": 2.3795867653469067e-07,
      "loss": 0.0001,
      "reward": 0.8326875865459442,
      "reward_std": 0.27240147814154625,
      "rewards/correct_code_reward_func": 0.35416667722165585,
      "rewards/len_reward_func": 0.47852087020874023,
      "step": 90
    },
    {
      "completion_length": 102.83333969116211,
      "epoch": 1.6017699115044248,
      "grad_norm": 22.765338134366964,
      "kl": 0.07025146484375,
      "learning_rate": 2.3314840961805802e-07,
      "loss": 0.0001,
      "reward": 0.686832070350647,
      "reward_std": 0.2984095811843872,
      "rewards/correct_code_reward_func": 0.2083333432674408,
      "rewards/len_reward_func": 0.47849874198436737,
      "step": 91
    },
    {
      "completion_length": 54.31250190734863,
      "epoch": 1.6194690265486726,
      "grad_norm": 1.6877966926685302,
      "kl": 0.0550537109375,
      "learning_rate": 2.2834440237932533e-07,
      "loss": 0.0001,
      "reward": 0.8750000298023224,
      "reward_std": 0.38857419788837433,
      "rewards/correct_code_reward_func": 0.4166666865348816,
      "rewards/len_reward_func": 0.4583333432674408,
      "step": 92
    },
    {
      "completion_length": 120.70833969116211,
      "epoch": 1.6371681415929205,
      "grad_norm": 10.817524146996126,
      "kl": 0.0389404296875,
      "learning_rate": 2.2354843931106932e-07,
      "loss": 0.0,
      "reward": 0.7042862474918365,
      "reward_std": 0.3615289777517319,
      "rewards/correct_code_reward_func": 0.2083333432674408,
      "rewards/len_reward_func": 0.49595288932323456,
      "step": 93
    },
    {
      "completion_length": 44.250000953674316,
      "epoch": 1.6548672566371683,
      "grad_norm": 28.86285310594792,
      "kl": 0.064697265625,
      "learning_rate": 2.1876230191778597e-07,
      "loss": 0.0001,
      "reward": 0.8750000298023224,
      "reward_std": 0.3857583850622177,
      "rewards/correct_code_reward_func": 0.3750000149011612,
      "rewards/len_reward_func": 0.5,
      "step": 94
    },
    {
      "completion_length": 53.62500190734863,
      "epoch": 1.672566371681416,
      "grad_norm": 3.9097872367150397,
      "kl": 0.0594482421875,
      "learning_rate": 2.1398776805413398e-07,
      "loss": 0.0001,
      "reward": 0.8750000298023224,
      "reward_std": 0.3233579397201538,
      "rewards/correct_code_reward_func": 0.3750000149011612,
      "rewards/len_reward_func": 0.5,
      "step": 95
    },
    {
      "completion_length": 65.00000190734863,
      "epoch": 1.6902654867256637,
      "grad_norm": 1.5650440517517894,
      "kl": 0.0361328125,
      "learning_rate": 2.092266112645343e-07,
      "loss": 0.0,
      "reward": 0.8589357733726501,
      "reward_std": 0.3421449810266495,
      "rewards/correct_code_reward_func": 0.3750000149011612,
      "rewards/len_reward_func": 0.48393575847148895,
      "step": 96
    },
    {
      "completion_length": 68.06250381469727,
      "epoch": 1.7079646017699115,
      "grad_norm": 2.91895596774148,
      "kl": 0.0269775390625,
      "learning_rate": 2.044806001243714e-07,
      "loss": 0.0,
      "reward": 0.9035088419914246,
      "reward_std": 0.48539453744888306,
      "rewards/correct_code_reward_func": 0.4375000298023224,
      "rewards/len_reward_func": 0.466008797287941,
      "step": 97
    },
    {
      "completion_length": 73.70833396911621,
      "epoch": 1.7256637168141593,
      "grad_norm": 24.67219025250006,
      "kl": 0.0465087890625,
      "learning_rate": 1.997514975830388e-07,
      "loss": 0.0,
      "reward": 0.8952526450157166,
      "reward_std": 0.29202311858534813,
      "rewards/correct_code_reward_func": 0.3958333432674408,
      "rewards/len_reward_func": 0.49941930174827576,
      "step": 98
    },
    {
      "completion_length": 97.75000381469727,
      "epoch": 1.7433628318584071,
      "grad_norm": 1.6430763507769786,
      "kl": 0.0406494140625,
      "learning_rate": 1.9504106030907603e-07,
      "loss": 0.0,
      "reward": 0.7537775337696075,
      "reward_std": 0.3211555927991867,
      "rewards/correct_code_reward_func": 0.2916666865348816,
      "rewards/len_reward_func": 0.46211084723472595,
      "step": 99
    },
    {
      "completion_length": 52.62500190734863,
      "epoch": 1.7610619469026547,
      "grad_norm": 3.9570608772498286,
      "kl": 0.0634765625,
      "learning_rate": 1.9035103803763792e-07,
      "loss": 0.0001,
      "reward": 0.8125,
      "reward_std": 0.3857453167438507,
      "rewards/correct_code_reward_func": 0.3125,
      "rewards/len_reward_func": 0.5,
      "step": 100
    },
    {
      "completion_length": 81.35416793823242,
      "epoch": 1.7787610619469025,
      "grad_norm": 2.731852842110385,
      "kl": 0.033447265625,
      "learning_rate": 1.8568317292053891e-07,
      "loss": 0.0,
      "reward": 0.9219505190849304,
      "reward_std": 0.2703188881278038,
      "rewards/correct_code_reward_func": 0.4375000149011612,
      "rewards/len_reward_func": 0.48445048928260803,
      "step": 101
    },
    {
      "completion_length": 83.31250381469727,
      "epoch": 1.7964601769911503,
      "grad_norm": 2.4613714386546595,
      "kl": 0.022613525390625,
      "learning_rate": 1.8103919887911523e-07,
      "loss": 0.0,
      "reward": 0.7291666865348816,
      "reward_std": 0.30859364569187164,
      "rewards/correct_code_reward_func": 0.2291666716337204,
      "rewards/len_reward_func": 0.5,
      "step": 102
    },
    {
      "completion_length": 79.33333587646484,
      "epoch": 1.8141592920353982,
      "grad_norm": 4.215262416117561,
      "kl": 0.040771484375,
      "learning_rate": 1.7642084096014402e-07,
      "loss": 0.0,
      "reward": 0.7859848439693451,
      "reward_std": 0.4127810299396515,
      "rewards/correct_code_reward_func": 0.291666679084301,
      "rewards/len_reward_func": 0.4943181872367859,
      "step": 103
    },
    {
      "completion_length": 58.52083396911621,
      "epoch": 1.831858407079646,
      "grad_norm": 4.745442984631991,
      "kl": 0.08740234375,
      "learning_rate": 1.7182981469505847e-07,
      "loss": 0.0001,
      "reward": 0.8333333730697632,
      "reward_std": 0.45660629868507385,
      "rewards/correct_code_reward_func": 0.354166679084301,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 104
    },
    {
      "completion_length": 112.64583587646484,
      "epoch": 1.8495575221238938,
      "grad_norm": 1.7418907068386247,
      "kl": 0.0325927734375,
      "learning_rate": 1.672678254626979e-07,
      "loss": 0.0,
      "reward": 0.6364757716655731,
      "reward_std": 0.3013785034418106,
      "rewards/correct_code_reward_func": 0.18750000558793545,
      "rewards/len_reward_func": 0.4489757716655731,
      "step": 105
    },
    {
      "completion_length": 57.5,
      "epoch": 1.8672566371681416,
      "grad_norm": 3.579303393703891,
      "kl": 0.0543212890625,
      "learning_rate": 1.6273656785582984e-07,
      "loss": 0.0001,
      "reward": 0.8536822199821472,
      "reward_std": 0.45756329596042633,
      "rewards/correct_code_reward_func": 0.3958333432674408,
      "rewards/len_reward_func": 0.45784883201122284,
      "step": 106
    },
    {
      "completion_length": 83.52083587646484,
      "epoch": 1.8849557522123894,
      "grad_norm": 2.1418835681881925,
      "kl": 0.0758056640625,
      "learning_rate": 1.5823772505167698e-07,
      "loss": 0.0001,
      "reward": 0.8958333432674408,
      "reward_std": 0.49633626639842987,
      "rewards/correct_code_reward_func": 0.4375000149011612,
      "rewards/len_reward_func": 0.4583333432674408,
      "step": 107
    },
    {
      "completion_length": 94.20833587646484,
      "epoch": 1.9026548672566372,
      "grad_norm": 32.45107688241668,
      "kl": 0.025146484375,
      "learning_rate": 1.5377296818668635e-07,
      "loss": 0.0,
      "reward": 0.854166716337204,
      "reward_std": 0.37177951633930206,
      "rewards/correct_code_reward_func": 0.3958333432674408,
      "rewards/len_reward_func": 0.4583333432674408,
      "step": 108
    },
    {
      "completion_length": 72.89583396911621,
      "epoch": 1.920353982300885,
      "grad_norm": 1.7762554154885124,
      "kl": 0.04364013671875,
      "learning_rate": 1.4934395573577013e-07,
      "loss": 0.0,
      "reward": 0.7892157137393951,
      "reward_std": 0.3116895519196987,
      "rewards/correct_code_reward_func": 0.3125000149011612,
      "rewards/len_reward_func": 0.47671568393707275,
      "step": 109
    },
    {
      "completion_length": 66.39583587646484,
      "epoch": 1.9380530973451329,
      "grad_norm": 1.3218901012489044,
      "kl": 0.02496337890625,
      "learning_rate": 1.4495233289624958e-07,
      "loss": 0.0,
      "reward": 0.7881554365158081,
      "reward_std": 0.3200981765985489,
      "rewards/correct_code_reward_func": 0.2916666716337204,
      "rewards/len_reward_func": 0.4964887797832489,
      "step": 110
    },
    {
      "completion_length": 58.291669845581055,
      "epoch": 1.9557522123893807,
      "grad_norm": 0.9195423141452592,
      "kl": 0.066162109375,
      "learning_rate": 1.4059973097673185e-07,
      "loss": 0.0001,
      "reward": 0.7083333432674408,
      "reward_std": 0.3382536321878433,
      "rewards/correct_code_reward_func": 0.2500000111758709,
      "rewards/len_reward_func": 0.4583333432674408,
      "step": 111
    },
    {
      "completion_length": 94.41667175292969,
      "epoch": 1.9734513274336283,
      "grad_norm": 89.52879883380525,
      "kl": 0.0601806640625,
      "learning_rate": 1.3628776679114514e-07,
      "loss": 0.0001,
      "reward": 0.7916666865348816,
      "reward_std": 0.31285394728183746,
      "rewards/correct_code_reward_func": 0.3125,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 112
    },
    {
      "completion_length": 100.27083587646484,
      "epoch": 1.991150442477876,
      "grad_norm": 2.390953296031483,
      "kl": 0.056396484375,
      "learning_rate": 1.320180420581587e-07,
      "loss": 0.0001,
      "reward": 0.6097994744777679,
      "reward_std": 0.18817735463380814,
      "rewards/correct_code_reward_func": 0.12500000558793545,
      "rewards/len_reward_func": 0.48479947447776794,
      "step": 113
    },
    {
      "completion_length": 63.875,
      "epoch": 2.0,
      "grad_norm": 3.601264581967262,
      "kl": 0.02001953125,
      "learning_rate": 1.2779214280620908e-07,
      "loss": 0.0,
      "reward": 0.6666666865348816,
      "reward_std": 0.17817416787147522,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.5,
      "step": 114
    },
    {
      "completion_length": 48.64583396911621,
      "epoch": 2.017699115044248,
      "grad_norm": 3.9192662388145028,
      "kl": 0.05908203125,
      "learning_rate": 1.2361163878435594e-07,
      "loss": 0.0001,
      "reward": 0.6176470816135406,
      "reward_std": 0.2277354598045349,
      "rewards/correct_code_reward_func": 0.12500000558793545,
      "rewards/len_reward_func": 0.49264705181121826,
      "step": 115
    },
    {
      "completion_length": 51.41666793823242,
      "epoch": 2.0353982300884956,
      "grad_norm": 2.2242727716589967,
      "kl": 0.04486083984375,
      "learning_rate": 1.1947808287918403e-07,
      "loss": 0.0,
      "reward": 0.7083333730697632,
      "reward_std": 0.2994871214032173,
      "rewards/correct_code_reward_func": 0.2083333395421505,
      "rewards/len_reward_func": 0.5,
      "step": 116
    },
    {
      "completion_length": 69.04166793823242,
      "epoch": 2.0530973451327434,
      "grad_norm": 5.407021499840827,
      "kl": 0.0469970703125,
      "learning_rate": 1.1539301053796949e-07,
      "loss": 0.0,
      "reward": 0.9791666865348816,
      "reward_std": 0.2946278229355812,
      "rewards/correct_code_reward_func": 0.4791666865348816,
      "rewards/len_reward_func": 0.5,
      "step": 117
    },
    {
      "completion_length": 47.83333396911621,
      "epoch": 2.0707964601769913,
      "grad_norm": 29.2301422488625,
      "kl": 0.08349609375,
      "learning_rate": 1.1135793919832334e-07,
      "loss": 0.0001,
      "reward": 0.9467871785163879,
      "reward_std": 0.43978360295295715,
      "rewards/correct_code_reward_func": 0.4791666716337204,
      "rewards/len_reward_func": 0.46762049198150635,
      "step": 118
    },
    {
      "completion_length": 90.41666793823242,
      "epoch": 2.088495575221239,
      "grad_norm": 1.804477992356325,
      "kl": 0.02587890625,
      "learning_rate": 1.0737436772452602e-07,
      "loss": 0.0,
      "reward": 0.7708333432674408,
      "reward_std": 0.3584126979112625,
      "rewards/correct_code_reward_func": 0.2708333358168602,
      "rewards/len_reward_func": 0.5,
      "step": 119
    },
    {
      "completion_length": 150.75000762939453,
      "epoch": 2.106194690265487,
      "grad_norm": 24.510631619520034,
      "kl": 0.0252685546875,
      "learning_rate": 1.0344377585075997e-07,
      "loss": 0.0,
      "reward": 0.684719979763031,
      "reward_std": 0.30992700159549713,
      "rewards/correct_code_reward_func": 0.1875000074505806,
      "rewards/len_reward_func": 0.4972199648618698,
      "step": 120
    },
    {
      "completion_length": 79.85416793823242,
      "epoch": 2.1238938053097347,
      "grad_norm": 2.667553033080654,
      "kl": 0.0318603515625,
      "learning_rate": 9.95676236314489e-08,
      "loss": 0.0,
      "reward": 0.8333333432674408,
      "reward_std": 0.31142252683639526,
      "rewards/correct_code_reward_func": 0.3333333432674408,
      "rewards/len_reward_func": 0.5,
      "step": 121
    },
    {
      "completion_length": 91.56250381469727,
      "epoch": 2.1415929203539825,
      "grad_norm": 100.44717449922435,
      "kl": 0.2076416015625,
      "learning_rate": 9.574735089890764e-08,
      "loss": 0.0002,
      "reward": 0.7708333432674408,
      "reward_std": 0.30231600999832153,
      "rewards/correct_code_reward_func": 0.2916666716337204,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 122
    },
    {
      "completion_length": 87.70833396911621,
      "epoch": 2.15929203539823,
      "grad_norm": 77.23064615733736,
      "kl": 0.22283935546875,
      "learning_rate": 9.198437672850248e-08,
      "loss": 0.0002,
      "reward": 0.770833358168602,
      "reward_std": 0.3205290399491787,
      "rewards/correct_code_reward_func": 0.2916666865348816,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 123
    },
    {
      "completion_length": 62.312503814697266,
      "epoch": 2.1769911504424777,
      "grad_norm": 17.734240204261667,
      "kl": 0.0535888671875,
      "learning_rate": 8.8280098911523e-08,
      "loss": 0.0001,
      "reward": 0.875,
      "reward_std": 0.2994871288537979,
      "rewards/correct_code_reward_func": 0.3958333432674408,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 124
    },
    {
      "completion_length": 65.41666793823242,
      "epoch": 2.1946902654867255,
      "grad_norm": 28.039058681720235,
      "kl": 0.0321044921875,
      "learning_rate": 8.463589343595976e-08,
      "loss": 0.0,
      "reward": 1.0833333730697632,
      "reward_std": 0.5049939155578613,
      "rewards/correct_code_reward_func": 0.5833333432674408,
      "rewards/len_reward_func": 0.5,
      "step": 125
    },
    {
      "completion_length": 98.64583396911621,
      "epoch": 2.2123893805309733,
      "grad_norm": 6.872794457305926,
      "kl": 0.0435791015625,
      "learning_rate": 8.105311397538084e-08,
      "loss": 0.0,
      "reward": 0.7077258378267288,
      "reward_std": 0.37921955436468124,
      "rewards/correct_code_reward_func": 0.25000000558793545,
      "rewards/len_reward_func": 0.4577258378267288,
      "step": 126
    },
    {
      "completion_length": 71.22916984558105,
      "epoch": 2.230088495575221,
      "grad_norm": 2.9985031046208857,
      "kl": 0.0357666015625,
      "learning_rate": 7.753309138609704e-08,
      "loss": 0.0,
      "reward": 0.7608647048473358,
      "reward_std": 0.35784420371055603,
      "rewards/correct_code_reward_func": 0.2708333358168602,
      "rewards/len_reward_func": 0.490031361579895,
      "step": 127
    },
    {
      "completion_length": 72.87500190734863,
      "epoch": 2.247787610619469,
      "grad_norm": 1.1881439057403123,
      "kl": 0.03509521484375,
      "learning_rate": 7.407713321280375e-08,
      "loss": 0.0,
      "reward": 0.854166716337204,
      "reward_std": 0.30231600999832153,
      "rewards/correct_code_reward_func": 0.354166679084301,
      "rewards/len_reward_func": 0.5,
      "step": 128
    },
    {
      "completion_length": 95.39583587646484,
      "epoch": 2.265486725663717,
      "grad_norm": 2.071026140631432,
      "kl": 0.0494384765625,
      "learning_rate": 7.06865232028808e-08,
      "loss": 0.0,
      "reward": 0.8162688612937927,
      "reward_std": 0.29165610671043396,
      "rewards/correct_code_reward_func": 0.3541666716337204,
      "rewards/len_reward_func": 0.46210217475891113,
      "step": 129
    },
    {
      "completion_length": 78.43750381469727,
      "epoch": 2.2831858407079646,
      "grad_norm": 0.7368071897691737,
      "kl": 0.027099609375,
      "learning_rate": 6.736252082953306e-08,
      "loss": 0.0,
      "reward": 0.8333333432674408,
      "reward_std": 0.2903675436973572,
      "rewards/correct_code_reward_func": 0.3333333358168602,
      "rewards/len_reward_func": 0.5,
      "step": 130
    },
    {
      "completion_length": 73.45833587646484,
      "epoch": 2.3008849557522124,
      "grad_norm": 2.1125816683606726,
      "kl": 0.0389404296875,
      "learning_rate": 6.410636082394771e-08,
      "loss": 0.0,
      "reward": 0.5384955406188965,
      "reward_std": 0.272975854575634,
      "rewards/correct_code_reward_func": 0.0833333358168602,
      "rewards/len_reward_func": 0.4551621973514557,
      "step": 131
    },
    {
      "completion_length": 101.02083587646484,
      "epoch": 2.3185840707964602,
      "grad_norm": 3.1749370092384415,
      "kl": 0.0380859375,
      "learning_rate": 6.091925271664156e-08,
      "loss": 0.0,
      "reward": 0.6666666716337204,
      "reward_std": 0.2903675250709057,
      "rewards/correct_code_reward_func": 0.1875,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 132
    },
    {
      "completion_length": 64.22916793823242,
      "epoch": 2.336283185840708,
      "grad_norm": 2.773648686745547,
      "kl": 0.11944580078125,
      "learning_rate": 5.7802380388170344e-08,
      "loss": 0.0001,
      "reward": 0.8125,
      "reward_std": 0.37034809589385986,
      "rewards/correct_code_reward_func": 0.3333333358168602,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 133
    },
    {
      "completion_length": 56.64583396911621,
      "epoch": 2.353982300884956,
      "grad_norm": 3.7650155223115793,
      "kl": 0.0516357421875,
      "learning_rate": 5.4756901629364886e-08,
      "loss": 0.0001,
      "reward": 1.129325270652771,
      "reward_std": 0.32795289158821106,
      "rewards/correct_code_reward_func": 0.6458333432674408,
      "rewards/len_reward_func": 0.4834919422864914,
      "step": 134
    },
    {
      "completion_length": 70.1875,
      "epoch": 2.3716814159292037,
      "grad_norm": 1.0135949886271995,
      "kl": 0.042236328125,
      "learning_rate": 5.178394771125968e-08,
      "loss": 0.0,
      "reward": 0.7218915224075317,
      "reward_std": 0.2874666675925255,
      "rewards/correct_code_reward_func": 0.2291666716337204,
      "rewards/len_reward_func": 0.49272486567497253,
      "step": 135
    },
    {
      "completion_length": 132.7916717529297,
      "epoch": 2.3893805309734515,
      "grad_norm": 1.3477438258822867,
      "kl": 0.03106689453125,
      "learning_rate": 4.888462296487128e-08,
      "loss": 0.0,
      "reward": 0.9166666865348816,
      "reward_std": 0.42218445241451263,
      "rewards/correct_code_reward_func": 0.4166666865348816,
      "rewards/len_reward_func": 0.5,
      "step": 136
    },
    {
      "completion_length": 60.22916793823242,
      "epoch": 2.4070796460176993,
      "grad_norm": 1.2650991231847002,
      "kl": 0.058837890625,
      "learning_rate": 4.606000437098476e-08,
      "loss": 0.0001,
      "reward": 0.8958333730697632,
      "reward_std": 0.2041093409061432,
      "rewards/correct_code_reward_func": 0.4166666716337204,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 137
    },
    {
      "completion_length": 64.6875,
      "epoch": 2.4247787610619467,
      "grad_norm": 2.901815806581681,
      "kl": 0.03662109375,
      "learning_rate": 4.3311141160099377e-08,
      "loss": 0.0,
      "reward": 0.9150572419166565,
      "reward_std": 0.3222523480653763,
      "rewards/correct_code_reward_func": 0.4166666716337204,
      "rewards/len_reward_func": 0.4983905553817749,
      "step": 138
    },
    {
      "completion_length": 92.79166984558105,
      "epoch": 2.442477876106195,
      "grad_norm": 0.6516445128630077,
      "kl": 0.03466796875,
      "learning_rate": 4.063905442268201e-08,
      "loss": 0.0,
      "reward": 0.6041666865348816,
      "reward_std": 0.16340987384319305,
      "rewards/correct_code_reward_func": 0.1041666679084301,
      "rewards/len_reward_func": 0.5,
      "step": 139
    },
    {
      "completion_length": 67.60416793823242,
      "epoch": 2.4601769911504423,
      "grad_norm": 1.8554639671156326,
      "kl": 0.0423583984375,
      "learning_rate": 3.8044736729873795e-08,
      "loss": 0.0,
      "reward": 0.9469756484031677,
      "reward_std": 0.43488872051239014,
      "rewards/correct_code_reward_func": 0.479166679084301,
      "rewards/len_reward_func": 0.46780893206596375,
      "step": 140
    },
    {
      "completion_length": 77.45833396911621,
      "epoch": 2.47787610619469,
      "grad_norm": 3.8300963552021754,
      "kl": 0.04541015625,
      "learning_rate": 3.552915176479071e-08,
      "loss": 0.0,
      "reward": 1.0416666865348816,
      "reward_std": 0.320542111992836,
      "rewards/correct_code_reward_func": 0.5416666716337204,
      "rewards/len_reward_func": 0.5,
      "step": 141
    },
    {
      "completion_length": 58.708335876464844,
      "epoch": 2.495575221238938,
      "grad_norm": 1.6137020616307847,
      "kl": 0.06591796875,
      "learning_rate": 3.309323396455446e-08,
      "loss": 0.0001,
      "reward": 0.9375000596046448,
      "reward_std": 0.2658637687563896,
      "rewards/correct_code_reward_func": 0.4583333432674408,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 142
    },
    {
      "completion_length": 86.12500381469727,
      "epoch": 2.5132743362831858,
      "grad_norm": 3.390075353900997,
      "kl": 0.0819091796875,
      "learning_rate": 3.073788817318707e-08,
      "loss": 0.0001,
      "reward": 0.8744058609008789,
      "reward_std": 0.429522842168808,
      "rewards/correct_code_reward_func": 0.3750000149011612,
      "rewards/len_reward_func": 0.4994058310985565,
      "step": 143
    },
    {
      "completion_length": 76.89583587646484,
      "epoch": 2.5309734513274336,
      "grad_norm": 17.278809046793295,
      "kl": 0.2576904296875,
      "learning_rate": 2.846398930549859e-08,
      "loss": 0.0003,
      "reward": 0.7083333432674408,
      "reward_std": 0.1178511306643486,
      "rewards/correct_code_reward_func": 0.20833333395421505,
      "rewards/len_reward_func": 0.5,
      "step": 144
    },
    {
      "completion_length": 86.625,
      "epoch": 2.5486725663716814,
      "grad_norm": 4.630135346925831,
      "kl": 0.0435791015625,
      "learning_rate": 2.62723820220917e-08,
      "loss": 0.0,
      "reward": 0.7428782880306244,
      "reward_std": 0.3244539424777031,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.4720449447631836,
      "step": 145
    },
    {
      "completion_length": 60.54166793823242,
      "epoch": 2.566371681415929,
      "grad_norm": 1.7403036344539455,
      "kl": 0.067626953125,
      "learning_rate": 2.416388041560491e-08,
      "loss": 0.0001,
      "reward": 1.0208333730697632,
      "reward_std": 0.24056155234575272,
      "rewards/correct_code_reward_func": 0.520833358168602,
      "rewards/len_reward_func": 0.5,
      "step": 146
    },
    {
      "completion_length": 72.68750190734863,
      "epoch": 2.584070796460177,
      "grad_norm": 8.835725978147865,
      "kl": 0.03240966796875,
      "learning_rate": 2.2139267708310454e-08,
      "loss": 0.0,
      "reward": 1.0000000596046448,
      "reward_std": 0.42218445241451263,
      "rewards/correct_code_reward_func": 0.5000000149011612,
      "rewards/len_reward_func": 0.5,
      "step": 147
    },
    {
      "completion_length": 65.31250381469727,
      "epoch": 2.601769911504425,
      "grad_norm": 21.996556093676393,
      "kl": 0.144287109375,
      "learning_rate": 2.019929596117889e-08,
      "loss": 0.0001,
      "reward": 0.9583333730697632,
      "reward_std": 0.4446708858013153,
      "rewards/correct_code_reward_func": 0.4583333432674408,
      "rewards/len_reward_func": 0.5,
      "step": 148
    },
    {
      "completion_length": 54.270835876464844,
      "epoch": 2.6194690265486726,
      "grad_norm": 4.1936490058839695,
      "kl": 0.099609375,
      "learning_rate": 1.8344685794519504e-08,
      "loss": 0.0001,
      "reward": 1.036210298538208,
      "reward_std": 0.37449972331523895,
      "rewards/correct_code_reward_func": 0.5416666716337204,
      "rewards/len_reward_func": 0.49454365670681,
      "step": 149
    },
    {
      "completion_length": 99.29166793823242,
      "epoch": 2.6371681415929205,
      "grad_norm": 63.75890524377285,
      "kl": 0.2841796875,
      "learning_rate": 1.6576126120299043e-08,
      "loss": 0.0003,
      "reward": 0.8559730350971222,
      "reward_std": 0.3255860209465027,
      "rewards/correct_code_reward_func": 0.3750000149011612,
      "rewards/len_reward_func": 0.4809730350971222,
      "step": 150
    },
    {
      "completion_length": 107.31250381469727,
      "epoch": 2.6548672566371683,
      "grad_norm": 36.89829026803691,
      "kl": 0.7239990234375,
      "learning_rate": 1.4894273886239206e-08,
      "loss": 0.0007,
      "reward": 0.6250000298023224,
      "reward_std": 0.3382536321878433,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.4583333432674408,
      "step": 151
    },
    {
      "completion_length": 64.35416793823242,
      "epoch": 2.672566371681416,
      "grad_norm": 2.096221752318607,
      "kl": 0.052490234375,
      "learning_rate": 1.3299753831787192e-08,
      "loss": 0.0001,
      "reward": 0.75,
      "reward_std": 0.38857419788837433,
      "rewards/correct_code_reward_func": 0.25,
      "rewards/len_reward_func": 0.5,
      "step": 152
    },
    {
      "completion_length": 96.64583587646484,
      "epoch": 2.6902654867256635,
      "grad_norm": 0.6573939843127445,
      "kl": 0.032470703125,
      "learning_rate": 1.1793158256050706e-08,
      "loss": 0.0,
      "reward": 0.7916666865348816,
      "reward_std": 0.3521219789981842,
      "rewards/correct_code_reward_func": 0.3125000074505806,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 153
    },
    {
      "completion_length": 57.45833396911621,
      "epoch": 2.7079646017699117,
      "grad_norm": 2.8347099035822936,
      "kl": 0.051513671875,
      "learning_rate": 1.0375046797782867e-08,
      "loss": 0.0001,
      "reward": 0.8958333730697632,
      "reward_std": 0.204109326004982,
      "rewards/correct_code_reward_func": 0.395833358168602,
      "rewards/len_reward_func": 0.5,
      "step": 154
    },
    {
      "completion_length": 97.02083587646484,
      "epoch": 2.725663716814159,
      "grad_norm": 89.9892277373503,
      "kl": 0.531982421875,
      "learning_rate": 9.045946227499297e-09,
      "loss": 0.0005,
      "reward": 0.9904971122741699,
      "reward_std": 0.4233323186635971,
      "rewards/correct_code_reward_func": 0.5208333432674408,
      "rewards/len_reward_func": 0.46966375410556793,
      "step": 155
    },
    {
      "completion_length": 75.37500381469727,
      "epoch": 2.7433628318584073,
      "grad_norm": 41.09603274193404,
      "kl": 0.0634765625,
      "learning_rate": 7.806350251804483e-09,
      "loss": 0.0001,
      "reward": 0.8541666865348816,
      "reward_std": 0.3492931127548218,
      "rewards/correct_code_reward_func": 0.3541666716337204,
      "rewards/len_reward_func": 0.5,
      "step": 156
    },
    {
      "completion_length": 92.85417175292969,
      "epoch": 2.7610619469026547,
      "grad_norm": 1.148928120100101,
      "kl": 0.0374755859375,
      "learning_rate": 6.6567193299996996e-09,
      "loss": 0.0,
      "reward": 0.7610294222831726,
      "reward_std": 0.485231876373291,
      "rewards/correct_code_reward_func": 0.3125,
      "rewards/len_reward_func": 0.4485294073820114,
      "step": 157
    },
    {
      "completion_length": 98.41667175292969,
      "epoch": 2.7787610619469025,
      "grad_norm": 0.2382852510613484,
      "kl": 0.02642822265625,
      "learning_rate": 5.597480503041485e-09,
      "loss": 0.0,
      "reward": 0.5000000149011612,
      "reward_std": 0.1178511306643486,
      "rewards/correct_code_reward_func": 0.0416666679084301,
      "rewards/len_reward_func": 0.4583333432674408,
      "step": 158
    },
    {
      "completion_length": 77.20833587646484,
      "epoch": 2.7964601769911503,
      "grad_norm": 6.517708531460766,
      "kl": 0.0968017578125,
      "learning_rate": 4.629027234912986e-09,
      "loss": 0.0001,
      "reward": 0.9473089575767517,
      "reward_std": 0.46372126042842865,
      "rewards/correct_code_reward_func": 0.5208333432674408,
      "rewards/len_reward_func": 0.4264756143093109,
      "step": 159
    },
    {
      "completion_length": 80.27083396911621,
      "epoch": 2.814159292035398,
      "grad_norm": 3.3787178462046454,
      "kl": 0.0364990234375,
      "learning_rate": 3.751719266468584e-09,
      "loss": 0.0,
      "reward": 0.8541666865348816,
      "reward_std": 0.13607724383473396,
      "rewards/correct_code_reward_func": 0.3541666679084301,
      "rewards/len_reward_func": 0.5,
      "step": 160
    },
    {
      "completion_length": 77.83333587646484,
      "epoch": 2.831858407079646,
      "grad_norm": 1.0432210255226626,
      "kl": 0.04864501953125,
      "learning_rate": 2.9658824818044325e-09,
      "loss": 0.0,
      "reward": 0.5659293830394745,
      "reward_std": 0.21801644191145897,
      "rewards/correct_code_reward_func": 0.08333333395421505,
      "rewards/len_reward_func": 0.4825960248708725,
      "step": 161
    },
    {
      "completion_length": 68.77083587646484,
      "epoch": 2.849557522123894,
      "grad_norm": 4.445236186593027,
      "kl": 0.061279296875,
      "learning_rate": 2.271808787206092e-09,
      "loss": 0.0001,
      "reward": 1.0243507027626038,
      "reward_std": 0.4195573776960373,
      "rewards/correct_code_reward_func": 0.6041666865348816,
      "rewards/len_reward_func": 0.4201839864253998,
      "step": 162
    },
    {
      "completion_length": 86.83333587646484,
      "epoch": 2.8672566371681416,
      "grad_norm": 4.1412440697628705,
      "kl": 0.5286865234375,
      "learning_rate": 1.6697560027171543e-09,
      "loss": 0.0005,
      "reward": 0.6875000149011612,
      "reward_std": 0.23144195601344109,
      "rewards/correct_code_reward_func": 0.2083333432674408,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 163
    },
    {
      "completion_length": 72.6875,
      "epoch": 2.8849557522123894,
      "grad_norm": 7.176170560573175,
      "kl": 0.04962158203125,
      "learning_rate": 1.1599477663696843e-09,
      "loss": 0.0,
      "reward": 0.7519232630729675,
      "reward_std": 0.26646677777171135,
      "rewards/correct_code_reward_func": 0.27083333395421505,
      "rewards/len_reward_func": 0.4810899347066879,
      "step": 164
    },
    {
      "completion_length": 87.64583587646484,
      "epoch": 2.9026548672566372,
      "grad_norm": 1.9523625673288243,
      "kl": 0.05029296875,
      "learning_rate": 7.425734511116998e-10,
      "loss": 0.0001,
      "reward": 0.6250000298023224,
      "reward_std": 0.22233545035123825,
      "rewards/correct_code_reward_func": 0.1250000037252903,
      "rewards/len_reward_func": 0.5,
      "step": 165
    },
    {
      "completion_length": 79.41666793823242,
      "epoch": 2.920353982300885,
      "grad_norm": 29.99743474150218,
      "kl": 0.346923828125,
      "learning_rate": 4.17788094463023e-10,
      "loss": 0.0003,
      "reward": 0.5346376597881317,
      "reward_std": 0.21488384157419205,
      "rewards/correct_code_reward_func": 0.06250000186264515,
      "rewards/len_reward_func": 0.4721376597881317,
      "step": 166
    },
    {
      "completion_length": 71.14583587646484,
      "epoch": 2.938053097345133,
      "grad_norm": 2.049279192654788,
      "kl": 0.037353515625,
      "learning_rate": 1.857123409250705e-10,
      "loss": 0.0,
      "reward": 0.8958333730697632,
      "reward_std": 0.24056155234575272,
      "rewards/correct_code_reward_func": 0.3958333432674408,
      "rewards/len_reward_func": 0.5,
      "step": 167
    },
    {
      "completion_length": 57.000003814697266,
      "epoch": 2.9557522123893807,
      "grad_norm": 11.198599869127536,
      "kl": 0.0416259765625,
      "learning_rate": 4.6432397166285e-11,
      "loss": 0.0,
      "reward": 0.9583333730697632,
      "reward_std": 0.29602527618408203,
      "rewards/correct_code_reward_func": 0.4791666716337204,
      "rewards/len_reward_func": 0.4791666716337204,
      "step": 168
    },
    {
      "completion_length": 50.625,
      "epoch": 2.9734513274336285,
      "grad_norm": 26.983307381671835,
      "kl": 0.271484375,
      "learning_rate": 0.0,
      "loss": 0.0003,
      "reward": 0.9375,
      "reward_std": 0.39486490190029144,
      "rewards/correct_code_reward_func": 0.4375,
      "rewards/len_reward_func": 0.5,
      "step": 169
    },
    {
      "epoch": 2.9734513274336285,
      "step": 169,
      "total_flos": 0.0,
      "train_loss": 6.70051409522318e-05,
      "train_runtime": 7347.2678,
      "train_samples_per_second": 0.138,
      "train_steps_per_second": 0.023
    }
  ],
  "logging_steps": 1,
  "max_steps": 169,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 50,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}