trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.04895960832313342,
  "eval_steps": 500,
  "global_step": 200,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.05,
      "learning_rate": 4.976714865090827e-05,
      "logits/chosen": -30.543472290039062,
      "logits/rejected": -67.6454086303711,
      "logps/chosen": -383.7469482421875,
      "logps/rejected": -404.8834228515625,
      "loss": 0.2274,
      "rewards/accuracies": 0.9768750071525574,
      "rewards/chosen": -33.849693298339844,
      "rewards/margins": 80.87075805664062,
      "rewards/rejected": -114.72045135498047,
      "step": 200
    }
  ],
  "logging_steps": 200,
  "max_steps": 2400,
  "num_train_epochs": 1,
  "save_steps": 200,
  "total_flos": 0.0,
  "trial_name": null,
  "trial_params": null
}