trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 1893,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.79,
      "learning_rate": 3.679344955097729e-05,
      "loss": 3.7932,
      "step": 500
    },
    {
      "epoch": 1.58,
      "learning_rate": 2.358689910195457e-05,
      "loss": 2.8486,
      "step": 1000
    },
    {
      "epoch": 2.38,
      "learning_rate": 1.0380348652931855e-05,
      "loss": 2.3623,
      "step": 1500
    },
    {
      "epoch": 3.0,
      "step": 1893,
      "total_flos": 539812211982336.0,
      "train_loss": 2.813352331936328,
      "train_runtime": 297.2362,
      "train_samples_per_second": 25.455,
      "train_steps_per_second": 6.369
    }
  ],
  "max_steps": 1893,
  "num_train_epochs": 3,
  "total_flos": 539812211982336.0,
  "trial_name": null,
  "trial_params": null
}