trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "global_step": 900,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.78,
      "learning_rate": 8.888888888888888e-06,
      "loss": 0.7681,
      "step": 500
    },
    {
      "epoch": 5.0,
      "step": 900,
      "total_flos": 1890764845013760.0,
      "train_loss": 0.529659177992079,
      "train_runtime": 599.6484,
      "train_samples_per_second": 47.936,
      "train_steps_per_second": 1.501
    }
  ],
  "max_steps": 900,
  "num_train_epochs": 5,
  "total_flos": 1890764845013760.0,
  "trial_name": null,
  "trial_params": null
}