trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "global_step": 8132,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "learning_rate": 5e-06,
      "loss": 2.0403,
      "step": 4066
    },
    {
      "epoch": 2.0,
      "learning_rate": 0.0,
      "loss": 0.394,
      "step": 8132
    },
    {
      "epoch": 2.0,
      "step": 8132,
      "total_flos": 1.0506418329673728e+16,
      "train_loss": 1.2171453878437424,
      "train_runtime": 2739.982,
      "train_samples_per_second": 23.742,
      "train_steps_per_second": 2.968
    }
  ],
  "max_steps": 8132,
  "num_train_epochs": 2,
  "total_flos": 1.0506418329673728e+16,
  "trial_name": null,
  "trial_params": null
}