trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.9473684210526314,
  "global_step": 18,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.95,
      "step": 18,
      "total_flos": 9.689634360285696e+16,
      "train_loss": 24.41952684190538,
      "train_runtime": 33.7077,
      "train_samples_per_second": 17.325,
      "train_steps_per_second": 0.534
    }
  ],
  "max_steps": 18,
  "num_train_epochs": 2,
  "total_flos": 9.689634360285696e+16,
  "trial_name": null,
  "trial_params": null
}