trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 6.0,
  "global_step": 846,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 3.55,
      "learning_rate": 8.17966903073286e-06,
      "loss": 0.0313,
      "step": 500
    },
    {
      "epoch": 6.0,
      "step": 846,
      "total_flos": 1418396775820992.0,
      "train_loss": 0.02046444483682619,
      "train_runtime": 257.6194,
      "train_samples_per_second": 52.403,
      "train_steps_per_second": 3.284
    }
  ],
  "max_steps": 846,
  "num_train_epochs": 6,
  "total_flos": 1418396775820992.0,
  "trial_name": null,
  "trial_params": null
}