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": -10.627127647399902,
      "logits/rejected": -7.845921516418457,
      "logps/chosen": -296.1811828613281,
      "logps/rejected": -265.0650634765625,
      "loss": 0.031,
      "rewards/accuracies": 0.9868749976158142,
      "rewards/chosen": 2.7860028743743896,
      "rewards/margins": 35.578575134277344,
      "rewards/rejected": -32.792572021484375,
      "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
}