{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.992,
  "eval_steps": 500,
  "global_step": 125,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "completion_length": 178.3541717529297,
      "epoch": 0.016,
      "grad_norm": 0.7053135404726241,
      "kl": 0.0,
      "learning_rate": 1.25e-07,
      "loss": 0.0,
      "reward": 0.5438157171010971,
      "reward_std": 0.4464171230792999,
      "rewards/correct_code_reward_func": 0.2500000074505806,
      "rewards/len_reward_func": 0.2938157171010971,
      "step": 1
    },
    {
      "completion_length": 176.89584350585938,
      "epoch": 0.032,
      "grad_norm": 2.3417507999605864,
      "kl": 0.0,
      "learning_rate": 2.5e-07,
      "loss": -0.0,
      "reward": 0.4928237199783325,
      "reward_std": 0.3440842181444168,
      "rewards/correct_code_reward_func": 0.1458333432674408,
      "rewards/len_reward_func": 0.34699034690856934,
      "step": 2
    },
    {
      "completion_length": 192.25,
      "epoch": 0.048,
      "grad_norm": 0.48071888167986226,
      "kl": -3.4809112548828125e-05,
      "learning_rate": 3.75e-07,
      "loss": -0.0,
      "reward": 0.449026882648468,
      "reward_std": 0.47666139900684357,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.2823602259159088,
      "step": 3
    },
    {
      "completion_length": 270.7291717529297,
      "epoch": 0.064,
      "grad_norm": 0.3569537292235904,
      "kl": 4.172325134277344e-07,
      "learning_rate": 5e-07,
      "loss": 0.0,
      "reward": 0.45451289415359497,
      "reward_std": 0.3296015188097954,
      "rewards/correct_code_reward_func": 0.1041666716337204,
      "rewards/len_reward_func": 0.3503462225198746,
      "step": 4
    },
    {
      "completion_length": 232.58334350585938,
      "epoch": 0.08,
      "grad_norm": 0.4439935569873824,
      "kl": -4.214048385620117e-05,
      "learning_rate": 4.999157413258781e-07,
      "loss": -0.0,
      "reward": 0.2809216380119324,
      "reward_std": 0.37323758006095886,
      "rewards/correct_code_reward_func": 0.0,
      "rewards/len_reward_func": 0.2809216380119324,
      "step": 5
    },
    {
      "completion_length": 261.9375,
      "epoch": 0.096,
      "grad_norm": 0.3541024748340305,
      "kl": -3.1054019927978516e-05,
      "learning_rate": 4.996630220997057e-07,
      "loss": -0.0,
      "reward": 0.4689294695854187,
      "reward_std": 0.4829525202512741,
      "rewards/correct_code_reward_func": 0.1875,
      "rewards/len_reward_func": 0.2814294695854187,
      "step": 6
    },
    {
      "completion_length": 306.12501525878906,
      "epoch": 0.112,
      "grad_norm": 0.3947267586718111,
      "kl": -2.968311309814453e-05,
      "learning_rate": 4.992420126717784e-07,
      "loss": -0.0,
      "reward": 0.3299577385187149,
      "reward_std": 0.30695800483226776,
      "rewards/correct_code_reward_func": 0.0,
      "rewards/len_reward_func": 0.3299577236175537,
      "step": 7
    },
    {
      "completion_length": 238.85417938232422,
      "epoch": 0.128,
      "grad_norm": 0.5118206169412205,
      "kl": -4.1604042053222656e-05,
      "learning_rate": 4.986529968316653e-07,
      "loss": 0.0,
      "reward": 0.4368641823530197,
      "reward_std": 0.2371533028781414,
      "rewards/correct_code_reward_func": 0.08333333395421505,
      "rewards/len_reward_func": 0.3535308539867401,
      "step": 8
    },
    {
      "completion_length": 304.7708435058594,
      "epoch": 0.144,
      "grad_norm": 0.3493288643723073,
      "kl": -5.060434341430664e-05,
      "learning_rate": 4.978963716169165e-07,
      "loss": -0.0,
      "reward": 0.4636567234992981,
      "reward_std": 0.46522316336631775,
      "rewards/correct_code_reward_func": 0.1666666679084301,
      "rewards/len_reward_func": 0.2969900518655777,
      "step": 9
    },
    {
      "completion_length": 264.4791717529297,
      "epoch": 0.16,
      "grad_norm": 0.8959331411324142,
      "kl": -7.329881191253662e-05,
      "learning_rate": 4.969726470454313e-07,
      "loss": -0.0,
      "reward": 0.5069085508584976,
      "reward_std": 0.37757958471775055,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.3402418941259384,
      "step": 10
    },
    {
      "completion_length": 201.3541717529297,
      "epoch": 0.176,
      "grad_norm": 0.4500851901199424,
      "kl": -4.0531158447265625e-05,
      "learning_rate": 4.958824457716706e-07,
      "loss": -0.0,
      "reward": 0.5010976195335388,
      "reward_std": 0.5051684230566025,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.334430992603302,
      "step": 11
    },
    {
      "completion_length": 297.37500762939453,
      "epoch": 0.192,
      "grad_norm": 0.43567877016900936,
      "kl": -4.696846008300781e-05,
      "learning_rate": 4.946265026669454e-07,
      "loss": -0.0,
      "reward": 0.3353367894887924,
      "reward_std": 0.3315615653991699,
      "rewards/correct_code_reward_func": 0.02083333395421505,
      "rewards/len_reward_func": 0.3145034611225128,
      "step": 12
    },
    {
      "completion_length": 201.75,
      "epoch": 0.208,
      "grad_norm": 0.5760425272270367,
      "kl": 2.5451183319091797e-05,
      "learning_rate": 4.932056643240618e-07,
      "loss": 0.0,
      "reward": 0.6330613493919373,
      "reward_std": 0.5388243198394775,
      "rewards/correct_code_reward_func": 0.3541666716337204,
      "rewards/len_reward_func": 0.27889466285705566,
      "step": 13
    },
    {
      "completion_length": 288.0416717529297,
      "epoch": 0.224,
      "grad_norm": 0.6576192766028296,
      "kl": 7.191300392150879e-05,
      "learning_rate": 4.916208884866592e-07,
      "loss": 0.0,
      "reward": 0.39246678352355957,
      "reward_std": 0.44181104004383087,
      "rewards/correct_code_reward_func": 0.08333333395421505,
      "rewards/len_reward_func": 0.3091334402561188,
      "step": 14
    },
    {
      "completion_length": 183.75000762939453,
      "epoch": 0.24,
      "grad_norm": 0.6134701851816485,
      "kl": 0.00014084577560424805,
      "learning_rate": 4.898732434036243e-07,
      "loss": 0.0,
      "reward": 0.42976467311382294,
      "reward_std": 0.3796156942844391,
      "rewards/correct_code_reward_func": 0.1250000037252903,
      "rewards/len_reward_func": 0.30476468801498413,
      "step": 15
    },
    {
      "completion_length": 207.77083587646484,
      "epoch": 0.256,
      "grad_norm": 0.7064422469677513,
      "kl": 0.000293731689453125,
      "learning_rate": 4.879639071090173e-07,
      "loss": 0.0,
      "reward": 0.35718174278736115,
      "reward_std": 0.408719003200531,
      "rewards/correct_code_reward_func": 0.0625,
      "rewards/len_reward_func": 0.29468175768852234,
      "step": 16
    },
    {
      "completion_length": 194.00000762939453,
      "epoch": 0.272,
      "grad_norm": 0.4827506793012771,
      "kl": 2.288818359375e-05,
      "learning_rate": 4.858941666279955e-07,
      "loss": -0.0,
      "reward": 0.5703845322132111,
      "reward_std": 0.39449170231819153,
      "rewards/correct_code_reward_func": 0.2500000111758709,
      "rewards/len_reward_func": 0.32038453221321106,
      "step": 17
    },
    {
      "completion_length": 238.18750762939453,
      "epoch": 0.288,
      "grad_norm": 0.41996129426983214,
      "kl": 0.00011217594146728516,
      "learning_rate": 4.836654171092682e-07,
      "loss": 0.0,
      "reward": 0.6033974885940552,
      "reward_std": 0.43378083407878876,
      "rewards/correct_code_reward_func": 0.2708333358168602,
      "rewards/len_reward_func": 0.33256417512893677,
      "step": 18
    },
    {
      "completion_length": 233.95834350585938,
      "epoch": 0.304,
      "grad_norm": 0.5123314162893023,
      "kl": 0.00013169646263122559,
      "learning_rate": 4.812791608846709e-07,
      "loss": 0.0,
      "reward": 0.40910618007183075,
      "reward_std": 0.3389303684234619,
      "rewards/correct_code_reward_func": 0.0833333358168602,
      "rewards/len_reward_func": 0.32577285170555115,
      "step": 19
    },
    {
      "completion_length": 316.3125,
      "epoch": 0.32,
      "grad_norm": 0.47220846544005757,
      "kl": 0.0002579689025878906,
      "learning_rate": 4.787370064564882e-07,
      "loss": 0.0,
      "reward": 0.41953830420970917,
      "reward_std": 0.4723764508962631,
      "rewards/correct_code_reward_func": 0.1458333358168602,
      "rewards/len_reward_func": 0.2737049460411072,
      "step": 20
    },
    {
      "completion_length": 227.7916717529297,
      "epoch": 0.336,
      "grad_norm": 0.6008188696097762,
      "kl": 0.0004267692565917969,
      "learning_rate": 4.7604066741321253e-07,
      "loss": 0.0,
      "reward": 0.6169633865356445,
      "reward_std": 0.48675188422203064,
      "rewards/correct_code_reward_func": 0.3333333432674408,
      "rewards/len_reward_func": 0.28363004326820374,
      "step": 21
    },
    {
      "completion_length": 252.27084350585938,
      "epoch": 0.352,
      "grad_norm": 0.5160855324353004,
      "kl": 0.0001907944679260254,
      "learning_rate": 4.731919612744659e-07,
      "loss": 0.0,
      "reward": 0.5648790299892426,
      "reward_std": 0.31752249598503113,
      "rewards/correct_code_reward_func": 0.18750000558793545,
      "rewards/len_reward_func": 0.37737900018692017,
      "step": 22
    },
    {
      "completion_length": 199.27083587646484,
      "epoch": 0.368,
      "grad_norm": 0.4296758598414565,
      "kl": 0.000273287296295166,
      "learning_rate": 4.7019280826586604e-07,
      "loss": 0.0,
      "reward": 0.45347318053245544,
      "reward_std": 0.3857118636369705,
      "rewards/correct_code_reward_func": 0.1041666716337204,
      "rewards/len_reward_func": 0.34930650889873505,
      "step": 23
    },
    {
      "completion_length": 221.77083587646484,
      "epoch": 0.384,
      "grad_norm": 1.2892124765727169,
      "kl": 0.0007638931274414062,
      "learning_rate": 4.6704523002466094e-07,
      "loss": 0.0,
      "reward": 0.4506646543741226,
      "reward_std": 0.20875498466193676,
      "rewards/correct_code_reward_func": 0.0416666679084301,
      "rewards/len_reward_func": 0.40899796783924103,
      "step": 24
    },
    {
      "completion_length": 266.8333435058594,
      "epoch": 0.4,
      "grad_norm": 1.4838265312357841,
      "kl": 0.0009393692016601562,
      "learning_rate": 4.6375134823700503e-07,
      "loss": 0.0,
      "reward": 0.4240592122077942,
      "reward_std": 0.36766907572746277,
      "rewards/correct_code_reward_func": 0.1041666679084301,
      "rewards/len_reward_func": 0.3198925405740738,
      "step": 25
    },
    {
      "completion_length": 225.20834350585938,
      "epoch": 0.416,
      "grad_norm": 0.45923128114507034,
      "kl": 0.0007772445678710938,
      "learning_rate": 4.603133832077953e-07,
      "loss": 0.0,
      "reward": 0.48156437277793884,
      "reward_std": 0.4315789043903351,
      "rewards/correct_code_reward_func": 0.1458333358168602,
      "rewards/len_reward_func": 0.33573102951049805,
      "step": 26
    },
    {
      "completion_length": 127.33333587646484,
      "epoch": 0.432,
      "grad_norm": 0.5853476018540942,
      "kl": 0.00115966796875,
      "learning_rate": 4.5673365236403216e-07,
      "loss": 0.0,
      "reward": 0.5744044184684753,
      "reward_std": 0.43071019649505615,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.30357107520103455,
      "step": 27
    },
    {
      "completion_length": 273.1666717529297,
      "epoch": 0.448,
      "grad_norm": 0.4925160720854453,
      "kl": 0.0010128021240234375,
      "learning_rate": 4.530145686927125e-07,
      "loss": 0.0,
      "reward": 0.5949445962905884,
      "reward_std": 0.42838945984840393,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.3241112679243088,
      "step": 28
    },
    {
      "completion_length": 178.75000762939453,
      "epoch": 0.464,
      "grad_norm": 1.1463419828713735,
      "kl": 0.0027008056640625,
      "learning_rate": 4.4915863911430897e-07,
      "loss": 0.0,
      "reward": 0.3950851857662201,
      "reward_std": 0.4124990254640579,
      "rewards/correct_code_reward_func": 0.08333333395421505,
      "rewards/len_reward_func": 0.3117518424987793,
      "step": 29
    },
    {
      "completion_length": 210.0,
      "epoch": 0.48,
      "grad_norm": 0.38472151178254205,
      "kl": 0.0011115074157714844,
      "learning_rate": 4.45168462792932e-07,
      "loss": 0.0,
      "reward": 0.48962171375751495,
      "reward_std": 0.443721666932106,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.32295504212379456,
      "step": 30
    },
    {
      "completion_length": 204.95833587646484,
      "epoch": 0.496,
      "grad_norm": 1.5507179471403478,
      "kl": 0.00183868408203125,
      "learning_rate": 4.4104672938431223e-07,
      "loss": 0.0,
      "reward": 0.6203064322471619,
      "reward_std": 0.39608660340309143,
      "rewards/correct_code_reward_func": 0.2916666865348816,
      "rewards/len_reward_func": 0.32863976061344147,
      "step": 31
    },
    {
      "completion_length": 159.83333587646484,
      "epoch": 0.512,
      "grad_norm": 0.9492121763047169,
      "kl": 0.001636505126953125,
      "learning_rate": 4.367962172227866e-07,
      "loss": 0.0,
      "reward": 0.5581120550632477,
      "reward_std": 0.4106949418783188,
      "rewards/correct_code_reward_func": 0.25,
      "rewards/len_reward_func": 0.3081120699644089,
      "step": 32
    },
    {
      "completion_length": 214.52083587646484,
      "epoch": 0.528,
      "grad_norm": 1.1197088117962448,
      "kl": 0.005927085876464844,
      "learning_rate": 4.324197914485075e-07,
      "loss": 0.0,
      "reward": 0.4833623319864273,
      "reward_std": 0.39532361924648285,
      "rewards/correct_code_reward_func": 0.1250000037252903,
      "rewards/len_reward_func": 0.3583623170852661,
      "step": 33
    },
    {
      "completion_length": 255.39583587646484,
      "epoch": 0.544,
      "grad_norm": 0.4856761942711223,
      "kl": 0.00038909912109375,
      "learning_rate": 4.2792040207614e-07,
      "loss": 0.0,
      "reward": 0.3419315367937088,
      "reward_std": 0.23230206966400146,
      "rewards/correct_code_reward_func": 0.02083333395421505,
      "rewards/len_reward_func": 0.3210982233285904,
      "step": 34
    },
    {
      "completion_length": 215.93750762939453,
      "epoch": 0.56,
      "grad_norm": 0.6317607844858494,
      "kl": 0.00643157958984375,
      "learning_rate": 4.2330108200634723e-07,
      "loss": 0.0,
      "reward": 0.48397234082221985,
      "reward_std": 0.36393553018569946,
      "rewards/correct_code_reward_func": 0.1458333432674408,
      "rewards/len_reward_func": 0.33813901245594025,
      "step": 35
    },
    {
      "completion_length": 170.52084350585938,
      "epoch": 0.576,
      "grad_norm": 0.47983643730367526,
      "kl": 0.0009899139404296875,
      "learning_rate": 4.185649449814045e-07,
      "loss": 0.0,
      "reward": 0.5446349233388901,
      "reward_std": 0.475721538066864,
      "rewards/correct_code_reward_func": 0.229166679084301,
      "rewards/len_reward_func": 0.3154682517051697,
      "step": 36
    },
    {
      "completion_length": 155.41666793823242,
      "epoch": 0.592,
      "grad_norm": 0.7182646382516522,
      "kl": 0.00705718994140625,
      "learning_rate": 4.137151834863213e-07,
      "loss": 0.0,
      "reward": 0.659955769777298,
      "reward_std": 0.4905228465795517,
      "rewards/correct_code_reward_func": 0.2916666865348816,
      "rewards/len_reward_func": 0.3682890981435776,
      "step": 37
    },
    {
      "completion_length": 234.35416793823242,
      "epoch": 0.608,
      "grad_norm": 0.5385402773601645,
      "kl": 0.001377105712890625,
      "learning_rate": 4.087550665968846e-07,
      "loss": 0.0,
      "reward": 0.34026579558849335,
      "reward_std": 0.42601510882377625,
      "rewards/correct_code_reward_func": 0.10416666977107525,
      "rewards/len_reward_func": 0.23609913140535355,
      "step": 38
    },
    {
      "completion_length": 150.02083587646484,
      "epoch": 0.624,
      "grad_norm": 0.8263933056203022,
      "kl": 0.0065155029296875,
      "learning_rate": 4.036879377760752e-07,
      "loss": 0.0,
      "reward": 0.6042845845222473,
      "reward_std": 0.4889480620622635,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.3334512561559677,
      "step": 39
    },
    {
      "completion_length": 168.77084350585938,
      "epoch": 0.64,
      "grad_norm": 0.6434934249206259,
      "kl": 0.00261688232421875,
      "learning_rate": 3.9851721262034157e-07,
      "loss": 0.0,
      "reward": 0.4757592976093292,
      "reward_std": 0.3624489903450012,
      "rewards/correct_code_reward_func": 0.125,
      "rewards/len_reward_func": 0.3507593274116516,
      "step": 40
    },
    {
      "completion_length": 194.0833396911621,
      "epoch": 0.656,
      "grad_norm": 0.5493289883401115,
      "kl": 0.003704071044921875,
      "learning_rate": 3.932463765572505e-07,
      "loss": 0.0,
      "reward": 0.42005764693021774,
      "reward_std": 0.37090964615345,
      "rewards/correct_code_reward_func": 0.16666667722165585,
      "rewards/len_reward_func": 0.25339096784591675,
      "step": 41
    },
    {
      "completion_length": 223.87500762939453,
      "epoch": 0.672,
      "grad_norm": 0.43766199341669654,
      "kl": 0.003650665283203125,
      "learning_rate": 3.8787898249606767e-07,
      "loss": 0.0,
      "reward": 0.35570722818374634,
      "reward_std": 0.3742265850305557,
      "rewards/correct_code_reward_func": 0.0416666679084301,
      "rewards/len_reward_func": 0.31404057145118713,
      "step": 42
    },
    {
      "completion_length": 159.93750762939453,
      "epoch": 0.688,
      "grad_norm": 0.662510453229864,
      "kl": 0.004608154296875,
      "learning_rate": 3.8241864843284964e-07,
      "loss": 0.0,
      "reward": 0.6946325600147247,
      "reward_std": 0.4132131338119507,
      "rewards/correct_code_reward_func": 0.3125000074505806,
      "rewards/len_reward_func": 0.38213254511356354,
      "step": 43
    },
    {
      "completion_length": 202.43750762939453,
      "epoch": 0.704,
      "grad_norm": 0.4902137587205668,
      "kl": 0.006561279296875,
      "learning_rate": 3.768690550116639e-07,
      "loss": 0.0,
      "reward": 0.4793711006641388,
      "reward_std": 0.4331624209880829,
      "rewards/correct_code_reward_func": 0.1875000074505806,
      "rewards/len_reward_func": 0.2918711006641388,
      "step": 44
    },
    {
      "completion_length": 201.52084350585938,
      "epoch": 0.72,
      "grad_norm": 1.5244716072983422,
      "kl": 0.0045623779296875,
      "learning_rate": 3.712339430435792e-07,
      "loss": 0.0,
      "reward": 0.376946821808815,
      "reward_std": 0.36263740062713623,
      "rewards/correct_code_reward_func": 0.0833333358168602,
      "rewards/len_reward_func": 0.2936134934425354,
      "step": 45
    },
    {
      "completion_length": 211.87501525878906,
      "epoch": 0.736,
      "grad_norm": 0.47019285013227063,
      "kl": 0.00263214111328125,
      "learning_rate": 3.65517110985099e-07,
      "loss": 0.0,
      "reward": 0.4974285364151001,
      "reward_std": 0.3727869838476181,
      "rewards/correct_code_reward_func": 0.12500000558793545,
      "rewards/len_reward_func": 0.3724285215139389,
      "step": 46
    },
    {
      "completion_length": 128.29166793823242,
      "epoch": 0.752,
      "grad_norm": 0.5038135341546401,
      "kl": 0.00347900390625,
      "learning_rate": 3.597224123777389e-07,
      "loss": 0.0,
      "reward": 0.608631819486618,
      "reward_std": 0.4802524596452713,
      "rewards/correct_code_reward_func": 0.3541666865348816,
      "rewards/len_reward_func": 0.25446511805057526,
      "step": 47
    },
    {
      "completion_length": 161.8541717529297,
      "epoch": 0.768,
      "grad_norm": 0.5836224104866172,
      "kl": 0.0087890625,
      "learning_rate": 3.5385375325047163e-07,
      "loss": 0.0,
      "reward": 0.5433830618858337,
      "reward_std": 0.4709310829639435,
      "rewards/correct_code_reward_func": 0.2291666679084301,
      "rewards/len_reward_func": 0.31421639025211334,
      "step": 48
    },
    {
      "completion_length": 140.47916793823242,
      "epoch": 0.784,
      "grad_norm": 0.49678584086470384,
      "kl": 0.0029144287109375,
      "learning_rate": 3.479150894867926e-07,
      "loss": 0.0,
      "reward": 0.5527551472187042,
      "reward_std": 0.5054349154233932,
      "rewards/correct_code_reward_func": 0.22916667722165585,
      "rewards/len_reward_func": 0.3235885202884674,
      "step": 49
    },
    {
      "completion_length": 190.87500762939453,
      "epoch": 0.8,
      "grad_norm": 1.1393985621399754,
      "kl": 0.004364013671875,
      "learning_rate": 3.4191042415818e-07,
      "loss": 0.0,
      "reward": 0.4087870866060257,
      "reward_std": 0.35092872381210327,
      "rewards/correct_code_reward_func": 0.1041666716337204,
      "rewards/len_reward_func": 0.3046204149723053,
      "step": 50
    },
    {
      "completion_length": 189.95833587646484,
      "epoch": 0.816,
      "grad_norm": 1.1299817912717918,
      "kl": 0.01027679443359375,
      "learning_rate": 3.3584380482574717e-07,
      "loss": 0.0,
      "reward": 0.48216497898101807,
      "reward_std": 0.35060346126556396,
      "rewards/correct_code_reward_func": 0.1875,
      "rewards/len_reward_func": 0.29466497898101807,
      "step": 51
    },
    {
      "completion_length": 155.20833587646484,
      "epoch": 0.832,
      "grad_norm": 1.0099136669800142,
      "kl": 0.0089263916015625,
      "learning_rate": 3.297193208119047e-07,
      "loss": 0.0,
      "reward": 0.7047297656536102,
      "reward_std": 0.537945419549942,
      "rewards/correct_code_reward_func": 0.3750000149011612,
      "rewards/len_reward_func": 0.32972970604896545,
      "step": 52
    },
    {
      "completion_length": 119.43750381469727,
      "epoch": 0.848,
      "grad_norm": 0.6910875224957795,
      "kl": 0.00537109375,
      "learning_rate": 3.235411004438741e-07,
      "loss": 0.0,
      "reward": 0.37493598461151123,
      "reward_std": 0.32232099026441574,
      "rewards/correct_code_reward_func": 0.0416666679084301,
      "rewards/len_reward_func": 0.33326931297779083,
      "step": 53
    },
    {
      "completion_length": 239.00001525878906,
      "epoch": 0.864,
      "grad_norm": 0.9366579226902805,
      "kl": 0.01202392578125,
      "learning_rate": 3.173133082709086e-07,
      "loss": 0.0,
      "reward": 0.5343351364135742,
      "reward_std": 0.39477548003196716,
      "rewards/correct_code_reward_func": 0.2083333358168602,
      "rewards/len_reward_func": 0.32600177824497223,
      "step": 54
    },
    {
      "completion_length": 165.6458396911621,
      "epoch": 0.88,
      "grad_norm": 0.5543105836651998,
      "kl": 0.0074310302734375,
      "learning_rate": 3.1104014225709784e-07,
      "loss": 0.0,
      "reward": 0.5656554698944092,
      "reward_std": 0.4120694398880005,
      "rewards/correct_code_reward_func": 0.2500000111758709,
      "rewards/len_reward_func": 0.315655454993248,
      "step": 55
    },
    {
      "completion_length": 224.8541717529297,
      "epoch": 0.896,
      "grad_norm": 0.4221238954900066,
      "kl": 0.0029735565185546875,
      "learning_rate": 3.0472583095164873e-07,
      "loss": 0.0,
      "reward": 0.4594677835702896,
      "reward_std": 0.3312383443117142,
      "rewards/correct_code_reward_func": 0.10416666977107525,
      "rewards/len_reward_func": 0.3553011566400528,
      "step": 56
    },
    {
      "completion_length": 135.8541717529297,
      "epoch": 0.912,
      "grad_norm": 0.7883647712663692,
      "kl": 0.005859375,
      "learning_rate": 2.983746306385499e-07,
      "loss": 0.0,
      "reward": 0.5004815310239792,
      "reward_std": 0.4576799273490906,
      "rewards/correct_code_reward_func": 0.1458333395421505,
      "rewards/len_reward_func": 0.3546481877565384,
      "step": 57
    },
    {
      "completion_length": 177.625,
      "epoch": 0.928,
      "grad_norm": 0.823591617621225,
      "kl": 0.00780487060546875,
      "learning_rate": 2.919908224675412e-07,
      "loss": 0.0,
      "reward": 0.47625844180583954,
      "reward_std": 0.4644129127264023,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.30959178507328033,
      "step": 58
    },
    {
      "completion_length": 111.95833587646484,
      "epoch": 0.944,
      "grad_norm": 0.7895834198702301,
      "kl": 0.01739501953125,
      "learning_rate": 2.8557870956832133e-07,
      "loss": 0.0,
      "reward": 0.45759040117263794,
      "reward_std": 0.33291806280612946,
      "rewards/correct_code_reward_func": 0.125,
      "rewards/len_reward_func": 0.3325904309749603,
      "step": 59
    },
    {
      "completion_length": 200.64583587646484,
      "epoch": 0.96,
      "grad_norm": 0.398008582749878,
      "kl": 0.00305938720703125,
      "learning_rate": 2.7914261414993976e-07,
      "loss": 0.0,
      "reward": 0.5845803320407867,
      "reward_std": 0.3288002014160156,
      "rewards/correct_code_reward_func": 0.2291666716337204,
      "rewards/len_reward_func": 0.35541366040706635,
      "step": 60
    },
    {
      "completion_length": 123.75000762939453,
      "epoch": 0.976,
      "grad_norm": 0.6258506341186686,
      "kl": 0.00853729248046875,
      "learning_rate": 2.726868745873286e-07,
      "loss": 0.0,
      "reward": 0.5053079277276993,
      "reward_std": 0.45688633620738983,
      "rewards/correct_code_reward_func": 0.16666667722165585,
      "rewards/len_reward_func": 0.3386412411928177,
      "step": 61
    },
    {
      "completion_length": 178.4166717529297,
      "epoch": 0.992,
      "grad_norm": 0.4226697044495165,
      "kl": 0.004604339599609375,
      "learning_rate": 2.662158424969357e-07,
      "loss": 0.0,
      "reward": 0.4181392341852188,
      "reward_std": 0.39034655690193176,
      "rewards/correct_code_reward_func": 0.1041666679084301,
      "rewards/len_reward_func": 0.3139725774526596,
      "step": 62
    },
    {
      "completion_length": 150.375,
      "epoch": 1.0,
      "grad_norm": 0.4226697044495165,
      "kl": 0.00823974609375,
      "learning_rate": 2.597338798034344e-07,
      "loss": 0.0,
      "reward": 0.7250348925590515,
      "reward_std": 0.3359350562095642,
      "rewards/correct_code_reward_func": 0.375,
      "rewards/len_reward_func": 0.3500348925590515,
      "step": 63
    },
    {
      "completion_length": 137.45833587646484,
      "epoch": 1.016,
      "grad_norm": 0.621781928712815,
      "kl": 0.0059356689453125,
      "learning_rate": 2.532453557994827e-07,
      "loss": 0.0,
      "reward": 0.5079408586025238,
      "reward_std": 0.43556541204452515,
      "rewards/correct_code_reward_func": 0.16666667722165585,
      "rewards/len_reward_func": 0.3412741720676422,
      "step": 64
    },
    {
      "completion_length": 237.45834350585938,
      "epoch": 1.032,
      "grad_norm": 0.4700296188876399,
      "kl": 0.00553131103515625,
      "learning_rate": 2.467546442005173e-07,
      "loss": 0.0,
      "reward": 0.3794639855623245,
      "reward_std": 0.3328210711479187,
      "rewards/correct_code_reward_func": 0.0625,
      "rewards/len_reward_func": 0.3169640153646469,
      "step": 65
    },
    {
      "completion_length": 217.33333587646484,
      "epoch": 1.048,
      "grad_norm": 0.47407157003975803,
      "kl": 0.01959228515625,
      "learning_rate": 2.4026612019656556e-07,
      "loss": 0.0,
      "reward": 0.5135317444801331,
      "reward_std": 0.3767416924238205,
      "rewards/correct_code_reward_func": 0.2083333395421505,
      "rewards/len_reward_func": 0.30519840121269226,
      "step": 66
    },
    {
      "completion_length": 159.9791717529297,
      "epoch": 1.064,
      "grad_norm": 1.33408581993753,
      "kl": 0.004638671875,
      "learning_rate": 2.337841575030642e-07,
      "loss": 0.0,
      "reward": 0.5043017268180847,
      "reward_std": 0.5010640621185303,
      "rewards/correct_code_reward_func": 0.1875000111758709,
      "rewards/len_reward_func": 0.3168017417192459,
      "step": 67
    },
    {
      "completion_length": 196.06250762939453,
      "epoch": 1.08,
      "grad_norm": 0.39338975551762256,
      "kl": 0.0045166015625,
      "learning_rate": 2.2731312541267143e-07,
      "loss": 0.0,
      "reward": 0.40096263587474823,
      "reward_std": 0.38369233906269073,
      "rewards/correct_code_reward_func": 0.1041666679084301,
      "rewards/len_reward_func": 0.29679596424102783,
      "step": 68
    },
    {
      "completion_length": 168.20833587646484,
      "epoch": 1.096,
      "grad_norm": 0.5733635012795789,
      "kl": 0.011474609375,
      "learning_rate": 2.2085738585006021e-07,
      "loss": 0.0,
      "reward": 0.5634751617908478,
      "reward_std": 0.42181093990802765,
      "rewards/correct_code_reward_func": 0.229166679084301,
      "rewards/len_reward_func": 0.3343084752559662,
      "step": 69
    },
    {
      "completion_length": 131.62500762939453,
      "epoch": 1.112,
      "grad_norm": 0.5768348851420809,
      "kl": 0.0180511474609375,
      "learning_rate": 2.1442129043167873e-07,
      "loss": 0.0,
      "reward": 0.44008754193782806,
      "reward_std": 0.4175649434328079,
      "rewards/correct_code_reward_func": 0.1458333358168602,
      "rewards/len_reward_func": 0.29425420612096786,
      "step": 70
    },
    {
      "completion_length": 178.1041717529297,
      "epoch": 1.1280000000000001,
      "grad_norm": 0.8562478784181595,
      "kl": 0.01507568359375,
      "learning_rate": 2.0800917753245875e-07,
      "loss": 0.0,
      "reward": 0.42609627544879913,
      "reward_std": 0.395632266998291,
      "rewards/correct_code_reward_func": 0.0833333358168602,
      "rewards/len_reward_func": 0.34276294708251953,
      "step": 71
    },
    {
      "completion_length": 142.43750381469727,
      "epoch": 1.144,
      "grad_norm": 0.5850086202761346,
      "kl": 0.007049560546875,
      "learning_rate": 2.0162536936145008e-07,
      "loss": 0.0,
      "reward": 0.6273844540119171,
      "reward_std": 0.37064287066459656,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.3565511107444763,
      "step": 72
    },
    {
      "completion_length": 173.4166717529297,
      "epoch": 1.16,
      "grad_norm": 1.0060077647295973,
      "kl": 0.01995849609375,
      "learning_rate": 1.9527416904835132e-07,
      "loss": 0.0,
      "reward": 0.6327731013298035,
      "reward_std": 0.40126484632492065,
      "rewards/correct_code_reward_func": 0.3125,
      "rewards/len_reward_func": 0.3202730864286423,
      "step": 73
    },
    {
      "completion_length": 196.5416717529297,
      "epoch": 1.176,
      "grad_norm": 0.5247248856581095,
      "kl": 0.004150390625,
      "learning_rate": 1.889598577429022e-07,
      "loss": 0.0,
      "reward": 0.4798154681921005,
      "reward_std": 0.4057523310184479,
      "rewards/correct_code_reward_func": 0.1458333395421505,
      "rewards/len_reward_func": 0.33398209512233734,
      "step": 74
    },
    {
      "completion_length": 163.37500762939453,
      "epoch": 1.192,
      "grad_norm": 0.44940299985092946,
      "kl": 0.00665283203125,
      "learning_rate": 1.8268669172909136e-07,
      "loss": 0.0,
      "reward": 0.6056158542633057,
      "reward_std": 0.4253086894750595,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.3347824960947037,
      "step": 75
    },
    {
      "completion_length": 127.60417175292969,
      "epoch": 1.208,
      "grad_norm": 0.5154656852815648,
      "kl": 0.0311279296875,
      "learning_rate": 1.7645889955612592e-07,
      "loss": 0.0,
      "reward": 0.5518045127391815,
      "reward_std": 0.46641653776168823,
      "rewards/correct_code_reward_func": 0.2083333358168602,
      "rewards/len_reward_func": 0.3434711843729019,
      "step": 76
    },
    {
      "completion_length": 126.47917175292969,
      "epoch": 1.224,
      "grad_norm": 0.7328993764471904,
      "kl": 0.01027679443359375,
      "learning_rate": 1.7028067918809535e-07,
      "loss": 0.0,
      "reward": 0.5090649425983429,
      "reward_std": 0.34914855659008026,
      "rewards/correct_code_reward_func": 0.20833333395421505,
      "rewards/len_reward_func": 0.3007315993309021,
      "step": 77
    },
    {
      "completion_length": 137.2291717529297,
      "epoch": 1.24,
      "grad_norm": 0.7418796199300344,
      "kl": 0.009124755859375,
      "learning_rate": 1.6415619517425294e-07,
      "loss": 0.0,
      "reward": 0.578598827123642,
      "reward_std": 0.4721776694059372,
      "rewards/correct_code_reward_func": 0.2500000074505806,
      "rewards/len_reward_func": 0.32859882712364197,
      "step": 78
    },
    {
      "completion_length": 158.95833587646484,
      "epoch": 1.256,
      "grad_norm": 2.068496131846233,
      "kl": 0.02191162109375,
      "learning_rate": 1.5808957584181994e-07,
      "loss": 0.0,
      "reward": 0.4869799315929413,
      "reward_std": 0.4040430933237076,
      "rewards/correct_code_reward_func": 0.1666666679084301,
      "rewards/len_reward_func": 0.3203132748603821,
      "step": 79
    },
    {
      "completion_length": 106.93750381469727,
      "epoch": 1.272,
      "grad_norm": 0.5762356445650323,
      "kl": 0.00872802734375,
      "learning_rate": 1.5208491051320744e-07,
      "loss": 0.0,
      "reward": 0.5087297856807709,
      "reward_std": 0.39580225944519043,
      "rewards/correct_code_reward_func": 0.1875000074505806,
      "rewards/len_reward_func": 0.3212297558784485,
      "step": 80
    },
    {
      "completion_length": 114.00000381469727,
      "epoch": 1.288,
      "grad_norm": 0.786665682978904,
      "kl": 0.013519287109375,
      "learning_rate": 1.461462467495284e-07,
      "loss": 0.0,
      "reward": 0.4894479066133499,
      "reward_std": 0.3821127265691757,
      "rewards/correct_code_reward_func": 0.125,
      "rewards/len_reward_func": 0.3644479066133499,
      "step": 81
    },
    {
      "completion_length": 154.83333587646484,
      "epoch": 1.304,
      "grad_norm": 0.7107423115065932,
      "kl": 0.0141754150390625,
      "learning_rate": 1.4027758762226107e-07,
      "loss": 0.0,
      "reward": 0.4811897426843643,
      "reward_std": 0.3424055427312851,
      "rewards/correct_code_reward_func": 0.1041666679084301,
      "rewards/len_reward_func": 0.3770230710506439,
      "step": 82
    },
    {
      "completion_length": 235.70834350585938,
      "epoch": 1.32,
      "grad_norm": 0.43515005998860457,
      "kl": 0.0047760009765625,
      "learning_rate": 1.3448288901490092e-07,
      "loss": 0.0,
      "reward": 0.4617680013179779,
      "reward_std": 0.3315645009279251,
      "rewards/correct_code_reward_func": 0.1250000037252903,
      "rewards/len_reward_func": 0.3367680013179779,
      "step": 83
    },
    {
      "completion_length": 143.1666717529297,
      "epoch": 1.336,
      "grad_norm": 0.65180392684865,
      "kl": 0.0086669921875,
      "learning_rate": 1.2876605695642084e-07,
      "loss": 0.0,
      "reward": 0.4678248018026352,
      "reward_std": 0.3193260580301285,
      "rewards/correct_code_reward_func": 0.1041666679084301,
      "rewards/len_reward_func": 0.3636581301689148,
      "step": 84
    },
    {
      "completion_length": 131.8333396911621,
      "epoch": 1.3519999999999999,
      "grad_norm": 0.4416459826072822,
      "kl": 0.01190185546875,
      "learning_rate": 1.231309449883361e-07,
      "loss": 0.0,
      "reward": 0.41201435029506683,
      "reward_std": 0.3407471626996994,
      "rewards/correct_code_reward_func": 0.0416666679084301,
      "rewards/len_reward_func": 0.37034766376018524,
      "step": 85
    },
    {
      "completion_length": 103.77083969116211,
      "epoch": 1.3679999999999999,
      "grad_norm": 0.43361682764693815,
      "kl": 0.0066986083984375,
      "learning_rate": 1.1758135156715041e-07,
      "loss": 0.0,
      "reward": 0.423152431845665,
      "reward_std": 0.3362526297569275,
      "rewards/correct_code_reward_func": 0.1041666679084301,
      "rewards/len_reward_func": 0.3189857602119446,
      "step": 86
    },
    {
      "completion_length": 176.375,
      "epoch": 1.384,
      "grad_norm": 1.352523990215239,
      "kl": 0.03180694580078125,
      "learning_rate": 1.1212101750393235e-07,
      "loss": 0.0,
      "reward": 0.4561140537261963,
      "reward_std": 0.39378371834754944,
      "rewards/correct_code_reward_func": 0.10416666977107525,
      "rewards/len_reward_func": 0.3519473969936371,
      "step": 87
    },
    {
      "completion_length": 79.97916793823242,
      "epoch": 1.4,
      "grad_norm": 2.141260004886994,
      "kl": 0.0297698974609375,
      "learning_rate": 1.0675362344274952e-07,
      "loss": 0.0,
      "reward": 0.42587006092071533,
      "reward_std": 0.4436161369085312,
      "rewards/correct_code_reward_func": 0.125,
      "rewards/len_reward_func": 0.30087001621723175,
      "step": 88
    },
    {
      "completion_length": 107.37500381469727,
      "epoch": 1.416,
      "grad_norm": 0.4125343610721507,
      "kl": 0.017974853515625,
      "learning_rate": 1.0148278737965844e-07,
      "loss": 0.0,
      "reward": 0.5323555767536163,
      "reward_std": 0.4073493778705597,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.36568886041641235,
      "step": 89
    },
    {
      "completion_length": 87.83333587646484,
      "epoch": 1.432,
      "grad_norm": 1.3465092335562199,
      "kl": 0.0328369140625,
      "learning_rate": 9.631206222392479e-08,
      "loss": 0.0,
      "reward": 0.5186317265033722,
      "reward_std": 0.4107673317193985,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.3519650846719742,
      "step": 90
    },
    {
      "completion_length": 124.33333969116211,
      "epoch": 1.448,
      "grad_norm": 0.7479671719828264,
      "kl": 0.02008056640625,
      "learning_rate": 9.124493340311537e-08,
      "loss": 0.0,
      "reward": 0.5086182951927185,
      "reward_std": 0.12347583472728729,
      "rewards/correct_code_reward_func": 0.0833333358168602,
      "rewards/len_reward_func": 0.4252849221229553,
      "step": 91
    },
    {
      "completion_length": 167.79166793823242,
      "epoch": 1.464,
      "grad_norm": 1.1543274217148969,
      "kl": 0.0083770751953125,
      "learning_rate": 8.628481651367875e-08,
      "loss": 0.0,
      "reward": 0.6109435856342316,
      "reward_std": 0.422005370259285,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.34011024236679077,
      "step": 92
    },
    {
      "completion_length": 115.33333969116211,
      "epoch": 1.48,
      "grad_norm": 0.3931440618018204,
      "kl": 0.0080108642578125,
      "learning_rate": 8.143505501859551e-08,
      "loss": 0.0,
      "reward": 0.44855794310569763,
      "reward_std": 0.32760028541088104,
      "rewards/correct_code_reward_func": 0.1250000037252903,
      "rewards/len_reward_func": 0.3235579580068588,
      "step": 93
    },
    {
      "completion_length": 151.39583587646484,
      "epoch": 1.496,
      "grad_norm": 0.5894110998499853,
      "kl": 0.009307861328125,
      "learning_rate": 7.669891799365282e-08,
      "loss": 0.0,
      "reward": 0.36864979565143585,
      "reward_std": 0.36151623725891113,
      "rewards/correct_code_reward_func": 0.0416666679084301,
      "rewards/len_reward_func": 0.32698309421539307,
      "step": 94
    },
    {
      "completion_length": 147.7916717529297,
      "epoch": 1.512,
      "grad_norm": 0.5972288037218009,
      "kl": 0.019500732421875,
      "learning_rate": 7.207959792385998e-08,
      "loss": 0.0,
      "reward": 0.4600509703159332,
      "reward_std": 0.4011112302541733,
      "rewards/correct_code_reward_func": 0.1041666679084301,
      "rewards/len_reward_func": 0.3558843284845352,
      "step": 95
    },
    {
      "completion_length": 149.9791717529297,
      "epoch": 1.528,
      "grad_norm": 0.86174582631064,
      "kl": 0.00836181640625,
      "learning_rate": 6.758020855149249e-08,
      "loss": 0.0,
      "reward": 0.44805125892162323,
      "reward_std": 0.33948560059070587,
      "rewards/correct_code_reward_func": 0.125,
      "rewards/len_reward_func": 0.32305125892162323,
      "step": 96
    },
    {
      "completion_length": 163.6041717529297,
      "epoch": 1.544,
      "grad_norm": 0.36337651628903067,
      "kl": 0.0181121826171875,
      "learning_rate": 6.320378277721342e-08,
      "loss": 0.0,
      "reward": 0.4679824113845825,
      "reward_std": 0.4093552529811859,
      "rewards/correct_code_reward_func": 0.10416666977107525,
      "rewards/len_reward_func": 0.3638157695531845,
      "step": 97
    },
    {
      "completion_length": 102.75000381469727,
      "epoch": 1.56,
      "grad_norm": 1.175706006208444,
      "kl": 0.023834228515625,
      "learning_rate": 5.895327061568775e-08,
      "loss": 0.0,
      "reward": 0.5549444258213043,
      "reward_std": 0.43229806423187256,
      "rewards/correct_code_reward_func": 0.2500000111758709,
      "rewards/len_reward_func": 0.3049444109201431,
      "step": 98
    },
    {
      "completion_length": 144.58333587646484,
      "epoch": 1.576,
      "grad_norm": 0.7970883391032116,
      "kl": 0.011474609375,
      "learning_rate": 5.483153720706798e-08,
      "loss": 0.0,
      "reward": 0.5082628130912781,
      "reward_std": 0.46417203545570374,
      "rewards/correct_code_reward_func": 0.2083333432674408,
      "rewards/len_reward_func": 0.2999294698238373,
      "step": 99
    },
    {
      "completion_length": 93.41667175292969,
      "epoch": 1.592,
      "grad_norm": 0.577298607650083,
      "kl": 0.0172119140625,
      "learning_rate": 5.0841360885690996e-08,
      "loss": 0.0,
      "reward": 0.43817608058452606,
      "reward_std": 0.4479677081108093,
      "rewards/correct_code_reward_func": 0.1458333358168602,
      "rewards/len_reward_func": 0.29234276711940765,
      "step": 100
    },
    {
      "completion_length": 112.79167175292969,
      "epoch": 1.608,
      "grad_norm": 0.6328737919939129,
      "kl": 0.0224151611328125,
      "learning_rate": 4.698543130728755e-08,
      "loss": 0.0,
      "reward": 0.6780606508255005,
      "reward_std": 0.37215377390384674,
      "rewards/correct_code_reward_func": 0.3333333358168602,
      "rewards/len_reward_func": 0.3447272926568985,
      "step": 101
    },
    {
      "completion_length": 125.33333587646484,
      "epoch": 1.624,
      "grad_norm": 0.8805964414937946,
      "kl": 0.007904052734375,
      "learning_rate": 4.326634763596784e-08,
      "loss": 0.0,
      "reward": 0.5231586992740631,
      "reward_std": 0.39199909567832947,
      "rewards/correct_code_reward_func": 0.125,
      "rewards/len_reward_func": 0.3981587141752243,
      "step": 102
    },
    {
      "completion_length": 143.81250762939453,
      "epoch": 1.6400000000000001,
      "grad_norm": 0.4566354927971138,
      "kl": 0.0088348388671875,
      "learning_rate": 3.968661679220467e-08,
      "loss": 0.0,
      "reward": 0.5204833149909973,
      "reward_std": 0.25895993411540985,
      "rewards/correct_code_reward_func": 0.1041666679084301,
      "rewards/len_reward_func": 0.4163166582584381,
      "step": 103
    },
    {
      "completion_length": 118.27083587646484,
      "epoch": 1.6560000000000001,
      "grad_norm": 0.41571642382372326,
      "kl": 0.00982666015625,
      "learning_rate": 3.624865176299499e-08,
      "loss": 0.0,
      "reward": 0.6309479027986526,
      "reward_std": 0.4369208961725235,
      "rewards/correct_code_reward_func": 0.29166667722165585,
      "rewards/len_reward_func": 0.33928124606609344,
      "step": 104
    },
    {
      "completion_length": 127.50000381469727,
      "epoch": 1.6720000000000002,
      "grad_norm": 0.44810915489413017,
      "kl": 0.021331787109375,
      "learning_rate": 3.295476997533905e-08,
      "loss": 0.0,
      "reward": 0.43531325459480286,
      "reward_std": 0.2511162757873535,
      "rewards/correct_code_reward_func": 0.0416666679084301,
      "rewards/len_reward_func": 0.39364662766456604,
      "step": 105
    },
    {
      "completion_length": 179.75000762939453,
      "epoch": 1.688,
      "grad_norm": 0.5235689630320457,
      "kl": 0.029052734375,
      "learning_rate": 2.980719173413396e-08,
      "loss": 0.0,
      "reward": 0.354349747300148,
      "reward_std": 0.2833855152130127,
      "rewards/correct_code_reward_func": 0.0416666679084301,
      "rewards/len_reward_func": 0.3126830607652664,
      "step": 106
    },
    {
      "completion_length": 120.50000381469727,
      "epoch": 1.704,
      "grad_norm": 1.2523814931919,
      "kl": 0.0085601806640625,
      "learning_rate": 2.680803872553408e-08,
      "loss": 0.0,
      "reward": 0.48751458525657654,
      "reward_std": 0.4157916307449341,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.32084792852401733,
      "step": 107
    },
    {
      "completion_length": 151.16666793823242,
      "epoch": 1.72,
      "grad_norm": 0.6447932351262547,
      "kl": 0.00946807861328125,
      "learning_rate": 2.395933258678745e-08,
      "loss": 0.0,
      "reward": 0.6216467022895813,
      "reward_std": 0.44435153901576996,
      "rewards/correct_code_reward_func": 0.2708333432674408,
      "rewards/len_reward_func": 0.3508133441209793,
      "step": 108
    },
    {
      "completion_length": 123.4375,
      "epoch": 1.736,
      "grad_norm": 0.5052351271825365,
      "kl": 0.0091552734375,
      "learning_rate": 2.1262993543511715e-08,
      "loss": 0.0,
      "reward": 0.6645375192165375,
      "reward_std": 0.4121406674385071,
      "rewards/correct_code_reward_func": 0.2916666716337204,
      "rewards/len_reward_func": 0.37287086248397827,
      "step": 109
    },
    {
      "completion_length": 153.47916793823242,
      "epoch": 1.752,
      "grad_norm": 0.6369640119331228,
      "kl": 0.0086212158203125,
      "learning_rate": 1.872083911532907e-08,
      "loss": 0.0,
      "reward": 0.40650297701358795,
      "reward_std": 0.27610746026039124,
      "rewards/correct_code_reward_func": 0.0416666679084301,
      "rewards/len_reward_func": 0.36483629047870636,
      "step": 110
    },
    {
      "completion_length": 132.1041717529297,
      "epoch": 1.768,
      "grad_norm": 0.44311314663042145,
      "kl": 0.01031494140625,
      "learning_rate": 1.6334582890731697e-08,
      "loss": 0.0,
      "reward": 0.7341299653053284,
      "reward_std": 0.504623532295227,
      "rewards/correct_code_reward_func": 0.4375,
      "rewards/len_reward_func": 0.29662999510765076,
      "step": 111
    },
    {
      "completion_length": 162.0625,
      "epoch": 1.784,
      "grad_norm": 0.6586562319299408,
      "kl": 0.0101318359375,
      "learning_rate": 1.4105833372004523e-08,
      "loss": 0.0,
      "reward": 0.4220695346593857,
      "reward_std": 0.2905489057302475,
      "rewards/correct_code_reward_func": 0.0416666679084301,
      "rewards/len_reward_func": 0.3804028630256653,
      "step": 112
    },
    {
      "completion_length": 106.89583587646484,
      "epoch": 1.8,
      "grad_norm": 0.7764527703399275,
      "kl": 0.019287109375,
      "learning_rate": 1.2036092890982619e-08,
      "loss": 0.0,
      "reward": 0.5292025506496429,
      "reward_std": 0.3711909055709839,
      "rewards/correct_code_reward_func": 0.1458333358168602,
      "rewards/len_reward_func": 0.38336920738220215,
      "step": 113
    },
    {
      "completion_length": 132.2916717529297,
      "epoch": 1.8159999999999998,
      "grad_norm": 0.6064692969361543,
      "kl": 0.009307861328125,
      "learning_rate": 1.0126756596375685e-08,
      "loss": 0.0,
      "reward": 0.48789724707603455,
      "reward_std": 0.3553258925676346,
      "rewards/correct_code_reward_func": 0.1458333358168602,
      "rewards/len_reward_func": 0.34206391870975494,
      "step": 114
    },
    {
      "completion_length": 120.22917175292969,
      "epoch": 1.8319999999999999,
      "grad_norm": 0.3843336433099292,
      "kl": 0.0174560546875,
      "learning_rate": 8.379111513340753e-09,
      "loss": 0.0,
      "reward": 0.47604209184646606,
      "reward_std": 0.3304741531610489,
      "rewards/correct_code_reward_func": 0.0833333358168602,
      "rewards/len_reward_func": 0.39270876348018646,
      "step": 115
    },
    {
      "completion_length": 145.31250762939453,
      "epoch": 1.8479999999999999,
      "grad_norm": 0.5532480666977985,
      "kl": 0.024169921875,
      "learning_rate": 6.7943356759381785e-09,
      "loss": 0.0,
      "reward": 0.46019650995731354,
      "reward_std": 0.3715391010046005,
      "rewards/correct_code_reward_func": 0.1041666679084301,
      "rewards/len_reward_func": 0.35602983832359314,
      "step": 116
    },
    {
      "completion_length": 149.93750381469727,
      "epoch": 1.8639999999999999,
      "grad_norm": 1.0372718715130624,
      "kl": 0.00921630859375,
      "learning_rate": 5.373497333054616e-09,
      "loss": 0.0,
      "reward": 0.4656513184309006,
      "reward_std": 0.44700081646442413,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.29898466169834137,
      "step": 117
    },
    {
      "completion_length": 144.89583587646484,
      "epoch": 1.88,
      "grad_norm": 0.4882633725665408,
      "kl": 0.014007568359375,
      "learning_rate": 4.117554228329406e-09,
      "loss": 0.0,
      "reward": 0.5118084400892258,
      "reward_std": 0.34739528596401215,
      "rewards/correct_code_reward_func": 0.1875000074505806,
      "rewards/len_reward_func": 0.32430844008922577,
      "step": 118
    },
    {
      "completion_length": 97.45833587646484,
      "epoch": 1.896,
      "grad_norm": 0.4820971241858704,
      "kl": 0.01251220703125,
      "learning_rate": 3.0273529545687125e-09,
      "loss": 0.0,
      "reward": 0.4883972406387329,
      "reward_std": 0.43138815462589264,
      "rewards/correct_code_reward_func": 0.12500000558793545,
      "rewards/len_reward_func": 0.3633972257375717,
      "step": 119
    },
    {
      "completion_length": 120.95833587646484,
      "epoch": 1.912,
      "grad_norm": 0.6000792918829387,
      "kl": 0.02130126953125,
      "learning_rate": 2.1036283830834224e-09,
      "loss": 0.0,
      "reward": 0.6599289178848267,
      "reward_std": 0.32202909141778946,
      "rewards/correct_code_reward_func": 0.2291666716337204,
      "rewards/len_reward_func": 0.4307622164487839,
      "step": 120
    },
    {
      "completion_length": 161.4791717529297,
      "epoch": 1.928,
      "grad_norm": 0.9160239570442854,
      "kl": 0.040740966796875,
      "learning_rate": 1.347003168334665e-09,
      "loss": 0.0,
      "reward": 0.6658598780632019,
      "reward_std": 0.45104770362377167,
      "rewards/correct_code_reward_func": 0.3750000149011612,
      "rewards/len_reward_func": 0.2908598631620407,
      "step": 121
    },
    {
      "completion_length": 197.33333587646484,
      "epoch": 1.944,
      "grad_norm": 1.2887565831293122,
      "kl": 0.0357666015625,
      "learning_rate": 7.579873282216598e-10,
      "loss": 0.0,
      "reward": 0.5509243905544281,
      "reward_std": 0.42525260150432587,
      "rewards/correct_code_reward_func": 0.229166679084301,
      "rewards/len_reward_func": 0.3217576891183853,
      "step": 122
    },
    {
      "completion_length": 238.39584350585938,
      "epoch": 1.96,
      "grad_norm": 0.36929428922486,
      "kl": 0.0052337646484375,
      "learning_rate": 3.3697790029424413e-10,
      "loss": 0.0,
      "reward": 0.5057644844055176,
      "reward_std": 0.4373111426830292,
      "rewards/correct_code_reward_func": 0.1458333395421505,
      "rewards/len_reward_func": 0.3599311411380768,
      "step": 123
    },
    {
      "completion_length": 126.41667175292969,
      "epoch": 1.976,
      "grad_norm": 0.4901359623088852,
      "kl": 0.007110595703125,
      "learning_rate": 8.425867412190091e-11,
      "loss": 0.0,
      "reward": 0.5390879511833191,
      "reward_std": 0.30444033443927765,
      "rewards/correct_code_reward_func": 0.1666666716337204,
      "rewards/len_reward_func": 0.3724212795495987,
      "step": 124
    },
    {
      "completion_length": 196.77083587646484,
      "epoch": 1.992,
      "grad_norm": 0.41890098929296765,
      "kl": 0.005767822265625,
      "learning_rate": 0.0,
      "loss": 0.0,
      "reward": 0.5094343423843384,
      "reward_std": 0.314863845705986,
      "rewards/correct_code_reward_func": 0.1041666679084301,
      "rewards/len_reward_func": 0.4052676856517792,
      "step": 125
    },
    {
      "epoch": 1.992,
      "step": 125,
      "total_flos": 0.0,
      "train_loss": 8.671089399598486e-06,
      "train_runtime": 8102.573,
      "train_samples_per_second": 0.093,
      "train_steps_per_second": 0.015
    }
  ],
  "logging_steps": 1,
  "max_steps": 125,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 25,
  "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
}