trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "global_step": 720,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.39,
      "learning_rate": 9.166666666666668e-06,
      "loss": 0.8341,
      "step": 500
    },
    {
      "epoch": 2.0,
      "step": 720,
      "total_flos": 966785769586176.0,
      "train_runtime": 502.9099,
      "train_samples_per_second": 22.863,
      "train_steps_per_second": 1.432
    }
  ],
  "max_steps": 720,
  "num_train_epochs": 2,
  "total_flos": 966785769586176.0,
  "trial_name": null,
  "trial_params": null
}