trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "global_step": 770,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 6.49,
      "learning_rate": 3.506493506493507e-08,
      "loss": 4.2363,
      "step": 500
    },
    {
      "epoch": 10.0,
      "step": 770,
      "total_flos": 6282413018234880.0,
      "train_runtime": 980.3735,
      "train_samples_per_second": 0.785
    }
  ],
  "max_steps": 770,
  "num_train_epochs": 10,
  "total_flos": 6282413018234880.0,
  "trial_name": null,
  "trial_params": null
}