trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "global_step": 1330,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 3.76,
      "learning_rate": 3.120300751879699e-05,
      "loss": 2.0422,
      "step": 500
    },
    {
      "epoch": 7.52,
      "learning_rate": 1.2406015037593984e-05,
      "loss": 0.93,
      "step": 1000
    },
    {
      "epoch": 10.0,
      "step": 1330,
      "total_flos": 2.999197549264896e+16,
      "train_loss": 1.282086261950041,
      "train_runtime": 3308.419,
      "train_samples_per_second": 20.082,
      "train_steps_per_second": 0.402
    }
  ],
  "max_steps": 1330,
  "num_train_epochs": 10,
  "total_flos": 2.999197549264896e+16,
  "trial_name": null,
  "trial_params": null
}