trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 1347,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.11,
      "learning_rate": 3.144023756495917e-05,
      "loss": 0.988,
      "step": 500
    },
    {
      "epoch": 2.23,
      "learning_rate": 1.2880475129918337e-05,
      "loss": 0.7995,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "step": 1347,
      "total_flos": 351574777921536.0,
      "train_loss": 0.8620750466894024,
      "train_runtime": 287.9872,
      "train_samples_per_second": 18.688,
      "train_steps_per_second": 4.677
    }
  ],
  "max_steps": 1347,
  "num_train_epochs": 3,
  "total_flos": 351574777921536.0,
  "trial_name": null,
  "trial_params": null
}