trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "global_step": 935,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.67,
      "learning_rate": 2.32620320855615e-05,
      "loss": 1.0352,
      "step": 500
    },
    {
      "epoch": 5.0,
      "step": 935,
      "total_flos": 1968088364605440.0,
      "train_loss": 1.0310750216723763,
      "train_runtime": 163.2592,
      "train_samples_per_second": 183.267,
      "train_steps_per_second": 5.727
    }
  ],
  "max_steps": 935,
  "num_train_epochs": 5,
  "total_flos": 1968088364605440.0,
  "trial_name": null,
  "trial_params": null
}