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": 1.229260935143288e-05,
      "loss": 1.3999,
      "step": 500
    },
    {
      "epoch": 3.0,
      "step": 663,
      "total_flos": 6291712793640960.0,
      "train_loss": 1.2881024851158913,
      "train_runtime": 953.9942,
      "train_samples_per_second": 1.387,
      "train_steps_per_second": 0.695
    }
  ],
  "max_steps": 663,
  "num_train_epochs": 3,
  "total_flos": 6291712793640960.0,
  "trial_name": null,
  "trial_params": null
}