trainer_state.json

{
  "best_metric": NaN,
  "best_model_checkpoint": "output/checkpoint-500",
  "epoch": 2.0833333333333335,
  "global_step": 500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.08,
      "learning_rate": 1.527777777777778e-05,
      "loss": 4.0735,
      "step": 500
    },
    {
      "epoch": 2.08,
      "eval_loss": NaN,
      "eval_runtime": 2.1474,
      "eval_samples_per_second": 223.064,
      "eval_steps_per_second": 111.765,
      "step": 500
    }
  ],
  "max_steps": 720,
  "num_train_epochs": 3,
  "total_flos": 12827512255488.0,
  "trial_name": null,
  "trial_params": null
}