{
  "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": 4.8242447519442067e-05,
      "learning_rate": 0.0029998415465061004,
      "loss": 0.0,
      "loss/policy_avg": 7.299240678548813e-08,
      "objective/entropy": 67.43842315673828,
      "objective/kl": 0.0,
      "objective/rlhf_reward": 1.5219240188598633,
      "objective/scores": 1.52203369140625,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 1,
      "timer/active_model_forward": 0.7594873309135437,
      "timer/calc_advantages": 31.33674430847168,
      "timer/calc_loss": 0.774461030960083,
      "timer/generate": 28.869869232177734,
      "timer/get_reward": 0.5197552442550659,
      "timer/model_forward": 0.7690438628196716,
      "timer/ref_model_forward": 0.7777600884437561,
      "timer/training_step": 34.545597076416016,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.0003168756065197156,
      "grad_norm": 0.00026670348597690463,
      "learning_rate": 0.002999683093012201,
      "loss": -0.0,
      "loss/policy_avg": -1.7043203115463257e-07,
      "objective/entropy": 118.22715759277344,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -3.824268341064453,
      "objective/scores": -3.82421875,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 2,
      "timer/active_model_forward": 0.6569761037826538,
      "timer/calc_advantages": 30.551788330078125,
      "timer/calc_loss": 0.6658673286437988,
      "timer/generate": 28.41728973388672,
      "timer/get_reward": 0.43157756328582764,
      "timer/model_forward": 0.6641299724578857,
      "timer/ref_model_forward": 0.67349773645401,
      "timer/training_step": 33.37031936645508,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.0004753134097795734,
      "grad_norm": 0.00013700829003937542,
      "learning_rate": 0.0029995246395183015,
      "loss": -0.0,
      "loss/policy_avg": -2.9243528842926025e-07,
      "objective/entropy": 1081.7860107421875,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -5.476104736328125,
      "objective/scores": -5.476806640625,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 3,
      "timer/active_model_forward": 0.9020705223083496,
      "timer/calc_advantages": 32.11817169189453,
      "timer/calc_loss": 0.9118615984916687,
      "timer/generate": 29.300296783447266,
      "timer/get_reward": 0.5959209203720093,
      "timer/model_forward": 0.9100804924964905,
      "timer/ref_model_forward": 0.9205952882766724,
      "timer/training_step": 35.83306121826172,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.0006337512130394313,
      "grad_norm": 7.692693907301873e-05,
      "learning_rate": 0.002999366186024402,
      "loss": -0.0,
      "loss/policy_avg": -2.0489096641540527e-07,
      "objective/entropy": 1181.2274169921875,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -5.0478363037109375,
      "objective/scores": -5.04833984375,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 4,
      "timer/active_model_forward": 0.8427364826202393,
      "timer/calc_advantages": 31.744182586669922,
      "timer/calc_loss": 0.8533791899681091,
      "timer/generate": 29.096416473388672,
      "timer/get_reward": 0.5531487464904785,
      "timer/model_forward": 0.851642906665802,
      "timer/ref_model_forward": 0.8617820143699646,
      "timer/training_step": 35.25868225097656,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.000792189016299289,
      "grad_norm": 3.911158273695037e-05,
      "learning_rate": 0.0029992077325305022,
      "loss": -0.0,
      "loss/policy_avg": -2.1513551473617554e-07,
      "objective/entropy": 1185.438232421875,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -4.198909759521484,
      "objective/scores": -4.198486328125,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 5,
      "timer/active_model_forward": 0.791987955570221,
      "timer/calc_advantages": 31.324127197265625,
      "timer/calc_loss": 0.8030194044113159,
      "timer/generate": 28.832427978515625,
      "timer/get_reward": 0.5202938914299011,
      "timer/model_forward": 0.8012962937355042,
      "timer/ref_model_forward": 0.8093616366386414,
      "timer/training_step": 34.65721130371094,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.0009506268195591468,
      "grad_norm": 2.5278823159169406e-05,
      "learning_rate": 0.002999049279036603,
      "loss": -0.0,
      "loss/policy_avg": -1.2619420886039734e-07,
      "objective/entropy": 528.8932495117188,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -3.1182708740234375,
      "objective/scores": -3.1180419921875,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 6,
      "timer/active_model_forward": 0.7996934652328491,
      "timer/calc_advantages": 31.30746078491211,
      "timer/calc_loss": 0.8107499480247498,
      "timer/generate": 28.788341522216797,
      "timer/get_reward": 0.5261460542678833,
      "timer/model_forward": 0.8090219497680664,
      "timer/ref_model_forward": 0.8171602487564087,
      "timer/training_step": 34.65579605102539,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.0011090646228190046,
      "grad_norm": 1.0207173772869282e-06,
      "learning_rate": 0.0029988908255427034,
      "loss": -0.0,
      "loss/policy_avg": -3.1781382858753204e-08,
      "objective/entropy": 1.5396817922592163,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -1.2192144393920898,
      "objective/scores": -1.2193603515625,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 7,
      "timer/active_model_forward": 1.0039308071136475,
      "timer/calc_advantages": 32.610069274902344,
      "timer/calc_loss": 1.0154139995574951,
      "timer/generate": 29.52051544189453,
      "timer/get_reward": 0.6604768633842468,
      "timer/model_forward": 1.0136762857437134,
      "timer/ref_model_forward": 1.0217972993850708,
      "timer/training_step": 36.71036911010742,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.0012675024260788625,
      "grad_norm": 1.4357982536239433e-06,
      "learning_rate": 0.0029987323720488037,
      "loss": -0.0,
      "loss/policy_avg": -7.380731403827667e-08,
      "objective/entropy": 2.126983404159546,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -1.3588581085205078,
      "objective/scores": -1.35894775390625,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 8,
      "timer/active_model_forward": 1.0782428979873657,
      "timer/calc_advantages": 33.10707092285156,
      "timer/calc_loss": 1.0924396514892578,
      "timer/generate": 29.794893264770508,
      "timer/get_reward": 0.7108417749404907,
      "timer/model_forward": 1.0906702280044556,
      "timer/ref_model_forward": 1.0941966772079468,
      "timer/training_step": 37.46529769897461,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.0014259402293387202,
      "grad_norm": 1.58626687607466e-06,
      "learning_rate": 0.002998573918554904,
      "loss": -0.0,
      "loss/policy_avg": -5.851325113326311e-08,
      "objective/entropy": 5.8357086181640625,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -1.3656387329101562,
      "objective/scores": -1.36566162109375,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 9,
      "timer/active_model_forward": 0.9783480167388916,
      "timer/calc_advantages": 32.70989990234375,
      "timer/calc_loss": 0.9892486333847046,
      "timer/generate": 29.658161163330078,
      "timer/get_reward": 0.6437245011329651,
      "timer/model_forward": 0.9874842166900635,
      "timer/ref_model_forward": 0.9964737892150879,
      "timer/training_step": 36.70344924926758,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.001584378032598578,
      "grad_norm": 1.8057213537758798e-06,
      "learning_rate": 0.002998415465061005,
      "loss": -0.0,
      "loss/policy_avg": -4.391768015921116e-08,
      "objective/entropy": 37.01044464111328,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -1.1269941329956055,
      "objective/scores": -1.126983642578125,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 10,
      "timer/active_model_forward": 0.7927595973014832,
      "timer/calc_advantages": 31.501224517822266,
      "timer/calc_loss": 0.8029042482376099,
      "timer/generate": 28.97046661376953,
      "timer/get_reward": 0.5210747718811035,
      "timer/model_forward": 0.8011468052864075,
      "timer/ref_model_forward": 0.8100461363792419,
      "timer/training_step": 34.83835983276367,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.0017428158358584358,
      "grad_norm": 1.3606485254058498e-06,
      "learning_rate": 0.002998257011567105,
      "loss": -0.0,
      "loss/policy_avg": -6.2282197177410126e-09,
      "objective/entropy": 82.50379943847656,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -0.20352506637573242,
      "objective/scores": -0.20351409912109375,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 11,
      "timer/active_model_forward": 0.8447611331939697,
      "timer/calc_advantages": 31.68113899230957,
      "timer/calc_loss": 0.8556842803955078,
      "timer/generate": 29.030838012695312,
      "timer/get_reward": 0.5556706786155701,
      "timer/model_forward": 0.853936493396759,
      "timer/ref_model_forward": 0.8623838424682617,
      "timer/training_step": 35.189205169677734,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.0019012536391182935,
      "grad_norm": 1.5168055824688054e-06,
      "learning_rate": 0.0029980985580732055,
      "loss": -0.0,
      "loss/policy_avg": -5.324545782059431e-08,
      "objective/entropy": 43.37924575805664,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -1.273842453956604,
      "objective/scores": -1.2738037109375,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 12,
      "timer/active_model_forward": 0.6771436929702759,
      "timer/calc_advantages": 30.566415786743164,
      "timer/calc_loss": 0.6872404217720032,
      "timer/generate": 28.381263732910156,
      "timer/get_reward": 0.4441335201263428,
      "timer/model_forward": 0.6854845881462097,
      "timer/ref_model_forward": 0.6946568489074707,
      "timer/training_step": 33.479454040527344,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.0020596914423781513,
      "grad_norm": 9.032095817929076e-07,
      "learning_rate": 0.0029979401045793063,
      "loss": -0.0,
      "loss/policy_avg": -5.2401446737349033e-08,
      "objective/entropy": 43.8469123840332,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -1.2306889295578003,
      "objective/scores": -1.230804443359375,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 13,
      "timer/active_model_forward": 0.8871802091598511,
      "timer/calc_advantages": 31.831775665283203,
      "timer/calc_loss": 0.8963775634765625,
      "timer/generate": 29.063201904296875,
      "timer/get_reward": 0.5812451839447021,
      "timer/model_forward": 0.8946273326873779,
      "timer/ref_model_forward": 0.9052593111991882,
      "timer/training_step": 35.49396514892578,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.002218129245638009,
      "grad_norm": 1.8163491404266097e-06,
      "learning_rate": 0.0029977816510854067,
      "loss": -0.0,
      "loss/policy_avg": -6.629852578043938e-08,
      "objective/entropy": 46.85642623901367,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -1.4318056106567383,
      "objective/scores": -1.431884765625,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 14,
      "timer/active_model_forward": 0.7645549774169922,
      "timer/calc_advantages": 31.06265640258789,
      "timer/calc_loss": 0.775362491607666,
      "timer/generate": 28.632740020751953,
      "timer/get_reward": 0.5032914876937866,
      "timer/model_forward": 0.7736347317695618,
      "timer/ref_model_forward": 0.7821637988090515,
      "timer/training_step": 34.282867431640625,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.002376567048897867,
      "grad_norm": 1.1021771797459223e-06,
      "learning_rate": 0.002997623197591507,
      "loss": -0.0,
      "loss/policy_avg": -4.4819898903369904e-08,
      "objective/entropy": 83.53251647949219,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -0.6715226173400879,
      "objective/scores": -0.671539306640625,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 15,
      "timer/active_model_forward": 1.1367318630218506,
      "timer/calc_advantages": 33.59859848022461,
      "timer/calc_loss": 1.2175476551055908,
      "timer/generate": 30.091690063476562,
      "timer/get_reward": 0.7473501563072205,
      "timer/model_forward": 1.2155828475952148,
      "timer/ref_model_forward": 1.1532230377197266,
      "timer/training_step": 38.26678466796875,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    },
    {
      "epoch": 0.002535004852157725,
      "grad_norm": 1.2331845482549397e-06,
      "learning_rate": 0.0029974647440976073,
      "loss": -0.0,
      "loss/policy_avg": -1.6909325495362282e-08,
      "objective/entropy": 119.47362518310547,
      "objective/kl": 0.0,
      "objective/rlhf_reward": -0.5125551223754883,
      "objective/scores": -0.5126724243164062,
      "policy/approxkl_avg": 0.0,
      "policy/clipfrac_avg": 0.0,
      "step": 16,
      "timer/active_model_forward": 0.9046459197998047,
      "timer/calc_advantages": 32.02373123168945,
      "timer/calc_loss": 0.914932906627655,
      "timer/generate": 29.190690994262695,
      "timer/get_reward": 0.5946121215820312,
      "timer/model_forward": 0.9131810665130615,
      "timer/ref_model_forward": 0.922136127948761,
      "timer/training_step": 35.75535202026367,
      "val/num_eos_tokens": 0.0,
      "val/ratio": 1.0
    }
  ],
  "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
}