trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 7.034327518289252,
  "global_step": 12500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.81,
      "learning_rate": 7.190315315315316e-05,
      "loss": 1.0104,
      "step": 5000
    },
    {
      "epoch": 5.63,
      "learning_rate": 4.375e-05,
      "loss": 0.7628,
      "step": 10000
    }
  ],
  "max_steps": 17770,
  "num_train_epochs": 10,
  "total_flos": 3.9182830534656e+16,
  "trial_name": null,
  "trial_params": null
}