trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 1008,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.49,
      "learning_rate": 2.5198412698412697e-05,
      "loss": 1.731,
      "step": 500
    },
    {
      "epoch": 2.98,
      "learning_rate": 3.9682539682539683e-07,
      "loss": 1.2292,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "step": 1008,
      "total_flos": 1826230648504320.0,
      "train_loss": 1.4770170771886433,
      "train_runtime": 362.6466,
      "train_samples_per_second": 27.796,
      "train_steps_per_second": 2.78
    }
  ],
  "max_steps": 1008,
  "num_train_epochs": 3,
  "total_flos": 1826230648504320.0,
  "trial_name": null,
  "trial_params": null
}