{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.0012675024260788625,
  "eval_steps": 500,
  "global_step": 8,
  "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
    }
  ],
  "logging_steps": 1,
  "max_steps": 18933,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 8,
  "total_flos": 4025485044940800.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}