trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.0,
  "global_step": 871,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.57,
      "learning_rate": 2.1297359357060852e-05,
      "loss": 3.5023,
      "step": 500
    },
    {
      "epoch": 1.0,
      "step": 871,
      "total_flos": 3639275421696000.0,
      "train_loss": 3.47868899007068,
      "train_runtime": 507.0283,
      "train_samples_per_second": 13.735,
      "train_steps_per_second": 1.718
    }
  ],
  "max_steps": 871,
  "num_train_epochs": 1,
  "total_flos": 3639275421696000.0,
  "trial_name": null,
  "trial_params": null
}