trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "eval_steps": 500,
  "global_step": 160,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.0,
      "step": 160,
      "total_flos": 760606430330880.0,
      "train_loss": 19.692240905761718,
      "train_runtime": 29.0912,
      "train_samples_per_second": 65.931,
      "train_steps_per_second": 5.5
    }
  ],
  "logging_steps": 500,
  "max_steps": 160,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 2,
  "save_steps": 500,
  "total_flos": 760606430330880.0,
  "train_batch_size": 12,
  "trial_name": null,
  "trial_params": null
}