{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.002535004852157725,
  "eval_steps": 500,
  "global_step": 16,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0001584378032598578,
      "grad_norm": 7.086519326549023e-05,
      "learning_rate": 2.9998415465061005e-05,
      "loss": 0.0,
      "loss/policy_avg": 9.534414857625961e-08,
      "objective/entropy": 66.7407455444336,
      "objective/kl": 0.0,
      "objective/rlhf_reward": 1.9394512176513672,
      "objective/scores": 1.9393310546875,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "policy/entropy_avg": 0.501089334487915,
      "step": 1,
      "timer/calc_advantages": 1.9189459085464478,
      "timer/calc_loss": 0.7234929800033569,
      "timer/get_reward": 0.5044295787811279,
      "timer/training_step": 5.130897045135498,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.0003168756065197156,
      "grad_norm": 9.20624828338623,
      "learning_rate": 2.999683093012201e-05,
      "loss": 0.0082,
      "loss/policy_avg": 0.008180337026715279,
      "objective/entropy": 58.23992919921875,
      "objective/kl": 0.04112038388848305,
      "objective/rlhf_reward": 2.586242914199829,
      "objective/scores": 2.59033203125,
      "policy/approxkl_avg": 0.13751985132694244,
      "policy/clipfrac_avg": 0.353515625,
      "policy/entropy_avg": 0.4371030628681183,
      "step": 2,
      "timer/calc_advantages": 1.77765953540802,
      "timer/calc_loss": 0.620478630065918,
      "timer/get_reward": 0.43380415439605713,
      "timer/training_step": 4.58284854888916,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0005027055740356,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.0004753134097795734,
      "grad_norm": 11.192896842956543,
      "learning_rate": 2.9995246395183014e-05,
      "loss": 0.0083,
      "loss/policy_avg": 0.008318130858242512,
      "objective/entropy": 60.747886657714844,
      "objective/kl": 0.1742033064365387,
      "objective/rlhf_reward": 2.7990851402282715,
      "objective/scores": 2.8162841796875,
      "policy/approxkl_avg": 0.1877279430627823,
      "policy/clipfrac_avg": 0.345703125,
      "policy/entropy_avg": 0.47007349133491516,
      "step": 3,
      "timer/calc_advantages": 2.159975528717041,
      "timer/calc_loss": 0.8658402562141418,
      "timer/get_reward": 0.5989301204681396,
      "timer/training_step": 5.865015029907227,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 0.9998154640197754,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.0006337512130394313,
      "grad_norm": 10.267037391662598,
      "learning_rate": 2.999366186024402e-05,
      "loss": 0.0119,
      "loss/policy_avg": 0.011887951754033566,
      "objective/entropy": 59.366294860839844,
      "objective/kl": 0.12045621126890182,
      "objective/rlhf_reward": 2.8014278411865234,
      "objective/scores": 2.8134765625,
      "policy/approxkl_avg": 0.2340196967124939,
      "policy/clipfrac_avg": 0.369140625,
      "policy/entropy_avg": 0.4660375118255615,
      "step": 4,
      "timer/calc_advantages": 1.9160572290420532,
      "timer/calc_loss": 0.8047055602073669,
      "timer/get_reward": 0.5554770231246948,
      "timer/training_step": 5.4731550216674805,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.000678539276123,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.000792189016299289,
      "grad_norm": 5.658022200805135e-05,
      "learning_rate": 2.9992077325305024e-05,
      "loss": 0.0,
      "loss/policy_avg": 9.505311027169228e-08,
      "objective/entropy": 63.056705474853516,
      "objective/kl": 0.0,
      "objective/rlhf_reward": 1.961860179901123,
      "objective/scores": 1.961883544921875,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "policy/entropy_avg": 0.49685803055763245,
      "step": 5,
      "timer/calc_advantages": 1.9551451206207275,
      "timer/calc_loss": 0.7570784091949463,
      "timer/get_reward": 0.5253068208694458,
      "timer/training_step": 5.274328231811523,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.0009506268195591468,
      "grad_norm": 5.630626201629639,
      "learning_rate": 2.9990492790366028e-05,
      "loss": 0.0036,
      "loss/policy_avg": 0.0035882075317204,
      "objective/entropy": 64.0039291381836,
      "objective/kl": -0.06272067129611969,
      "objective/rlhf_reward": 2.5381531715393066,
      "objective/scores": 2.53204345703125,
      "policy/approxkl_avg": 0.07097644358873367,
      "policy/clipfrac_avg": 0.244140625,
      "policy/entropy_avg": 0.49359244108200073,
      "step": 6,
      "timer/calc_advantages": 1.9771534204483032,
      "timer/calc_loss": 0.7805941700935364,
      "timer/get_reward": 0.5281075835227966,
      "timer/training_step": 5.320864677429199,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.000971794128418,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.0011090646228190046,
      "grad_norm": 7.572779178619385,
      "learning_rate": 2.9988908255427033e-05,
      "loss": 0.0058,
      "loss/policy_avg": 0.005815813317894936,
      "objective/entropy": 69.00190734863281,
      "objective/kl": 0.07886971533298492,
      "objective/rlhf_reward": 3.3839988708496094,
      "objective/scores": 3.39208984375,
      "policy/approxkl_avg": 0.1876513808965683,
      "policy/clipfrac_avg": 0.34765625,
      "policy/entropy_avg": 0.520235538482666,
      "step": 7,
      "timer/calc_advantages": 2.296631336212158,
      "timer/calc_loss": 0.9676017761230469,
      "timer/get_reward": 0.6649996042251587,
      "timer/training_step": 6.399840831756592,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.000232458114624,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.0012675024260788625,
      "grad_norm": 11.984278678894043,
      "learning_rate": 2.9987323720488037e-05,
      "loss": 0.0093,
      "loss/policy_avg": 0.009322328492999077,
      "objective/entropy": 66.09063720703125,
      "objective/kl": 0.015054229646921158,
      "objective/rlhf_reward": 2.489471673965454,
      "objective/scores": 2.491485595703125,
      "policy/approxkl_avg": 0.18275578320026398,
      "policy/clipfrac_avg": 0.3046875,
      "policy/entropy_avg": 0.5070106387138367,
      "step": 8,
      "timer/calc_advantages": 2.2399752140045166,
      "timer/calc_loss": 1.0477569103240967,
      "timer/get_reward": 0.7243468761444092,
      "timer/training_step": 6.686029434204102,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0009795427322388,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.0014259402293387202,
      "grad_norm": 7.540817750850692e-05,
      "learning_rate": 2.9985739185549042e-05,
      "loss": 0.0,
      "loss/policy_avg": 1.367880031466484e-07,
      "objective/entropy": 58.587005615234375,
      "objective/kl": 0.0,
      "objective/rlhf_reward": 2.611638069152832,
      "objective/scores": 2.61175537109375,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "policy/entropy_avg": 0.42249488830566406,
      "step": 9,
      "timer/calc_advantages": 2.2680113315582275,
      "timer/calc_loss": 0.9395630359649658,
      "timer/get_reward": 0.6489749550819397,
      "timer/training_step": 6.220600128173828,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.001584378032598578,
      "grad_norm": 1.935127854347229,
      "learning_rate": 2.9984154650610046e-05,
      "loss": 0.0007,
      "loss/policy_avg": 0.0006855675601400435,
      "objective/entropy": 55.21089172363281,
      "objective/kl": 0.0106657724827528,
      "objective/rlhf_reward": 2.033531427383423,
      "objective/scores": 2.0347900390625,
      "policy/approxkl_avg": 0.02970672771334648,
      "policy/clipfrac_avg": 0.197265625,
      "policy/entropy_avg": 0.4399160146713257,
      "step": 10,
      "timer/calc_advantages": 1.9682772159576416,
      "timer/calc_loss": 0.7604256272315979,
      "timer/get_reward": 0.5244199633598328,
      "timer/training_step": 5.279101371765137,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0002206563949585,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.0017428158358584358,
      "grad_norm": 2.560812473297119,
      "learning_rate": 2.998257011567105e-05,
      "loss": 0.0011,
      "loss/policy_avg": 0.0011356198228895664,
      "objective/entropy": 64.66464233398438,
      "objective/kl": 0.009169694036245346,
      "objective/rlhf_reward": 2.824911594390869,
      "objective/scores": 2.82574462890625,
      "policy/approxkl_avg": 0.03749970346689224,
      "policy/clipfrac_avg": 0.21484375,
      "policy/entropy_avg": 0.5036777257919312,
      "step": 11,
      "timer/calc_advantages": 2.0788302421569824,
      "timer/calc_loss": 0.8143972754478455,
      "timer/get_reward": 0.5581433176994324,
      "timer/training_step": 5.590597629547119,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.000451683998108,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.0019012536391182935,
      "grad_norm": 4.297569751739502,
      "learning_rate": 2.9980985580732055e-05,
      "loss": 0.0034,
      "loss/policy_avg": 0.003396428655833006,
      "objective/entropy": 69.60541534423828,
      "objective/kl": 0.10166570544242859,
      "objective/rlhf_reward": 1.655887246131897,
      "objective/scores": 1.6659393310546875,
      "policy/approxkl_avg": 0.09505846351385117,
      "policy/clipfrac_avg": 0.318359375,
      "policy/entropy_avg": 0.534148633480072,
      "step": 12,
      "timer/calc_advantages": 1.8084050416946411,
      "timer/calc_loss": 0.6441062092781067,
      "timer/get_reward": 0.44651317596435547,
      "timer/training_step": 4.796171188354492,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 0.9999833106994629,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.0020596914423781513,
      "grad_norm": 7.657416426809505e-05,
      "learning_rate": 2.997940104579306e-05,
      "loss": 0.0,
      "loss/policy_avg": 1.4971010386943817e-07,
      "objective/entropy": 58.742462158203125,
      "objective/kl": 0.0,
      "objective/rlhf_reward": 3.092315673828125,
      "objective/scores": 3.0924072265625,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "policy/entropy_avg": 0.4520529806613922,
      "step": 13,
      "timer/calc_advantages": 2.120788335800171,
      "timer/calc_loss": 0.8547518849372864,
      "timer/get_reward": 0.5879230499267578,
      "timer/training_step": 5.762821197509766,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.002218129245638009,
      "grad_norm": 2.661646604537964,
      "learning_rate": 2.9977816510854064e-05,
      "loss": 0.001,
      "loss/policy_avg": 0.001029975712299347,
      "objective/entropy": 68.30726623535156,
      "objective/kl": 0.03951136767864227,
      "objective/rlhf_reward": 2.7953004837036133,
      "objective/scores": 2.7994384765625,
      "policy/approxkl_avg": 0.040699079632759094,
      "policy/clipfrac_avg": 0.2265625,
      "policy/entropy_avg": 0.5269634127616882,
      "step": 14,
      "timer/calc_advantages": 1.9552991390228271,
      "timer/calc_loss": 0.7335948944091797,
      "timer/get_reward": 0.5061210989952087,
      "timer/training_step": 5.143916130065918,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.000035047531128,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.002376567048897867,
      "grad_norm": 3.1387367248535156,
      "learning_rate": 2.997623197591507e-05,
      "loss": 0.0025,
      "loss/policy_avg": 0.00247659208253026,
      "objective/entropy": 56.77816390991211,
      "objective/kl": 0.006408168934285641,
      "objective/rlhf_reward": 2.673807144165039,
      "objective/scores": 2.6744384765625,
      "policy/approxkl_avg": 0.06710580736398697,
      "policy/clipfrac_avg": 0.283203125,
      "policy/entropy_avg": 0.44521695375442505,
      "step": 15,
      "timer/calc_advantages": 2.5635008811950684,
      "timer/calc_loss": 1.1095852851867676,
      "timer/get_reward": 0.7531031966209412,
      "timer/training_step": 7.173662185668945,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0003986358642578,
      "val/ratio_var": NaN
    },
    {
      "epoch": 0.002535004852157725,
      "grad_norm": 3.675922155380249,
      "learning_rate": 2.9974647440976074e-05,
      "loss": 0.0035,
      "loss/policy_avg": 0.0034786476753652096,
      "objective/entropy": 55.65203094482422,
      "objective/kl": 0.0999540314078331,
      "objective/rlhf_reward": 2.479569435119629,
      "objective/scores": 2.489501953125,
      "policy/approxkl_avg": 0.10638778656721115,
      "policy/clipfrac_avg": 0.359375,
      "policy/entropy_avg": 0.4279758334159851,
      "step": 16,
      "timer/calc_advantages": 2.1651806831359863,
      "timer/calc_loss": 0.8663696646690369,
      "timer/get_reward": 0.5983204245567322,
      "timer/training_step": 5.955389022827148,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 0.9999206066131592,
      "val/ratio_var": NaN
    }
  ],
  "logging_steps": 1,
  "max_steps": 18933,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 8,
  "total_flos": 8202719718014976.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}