trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 663,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.26,
      "learning_rate": 5.101088646967341e-06,
      "loss": 0.0571,
      "step": 500
    },
    {
      "epoch": 3.0,
      "step": 663,
      "total_flos": 12409540182016.0,
      "train_loss": 0.04392106060169042,
      "train_runtime": 53436.2788,
      "train_samples_per_second": 1.191,
      "train_steps_per_second": 0.012
    }
  ],
  "max_steps": 663,
  "num_train_epochs": 3,
  "total_flos": 12409540182016.0,
  "trial_name": null,
  "trial_params": null
}