trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 18.51851851851852,
  "eval_steps": 500,
  "global_step": 1500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 6.17,
      "learning_rate": 3.45679012345679e-05,
      "loss": 2.7366,
      "step": 500
    },
    {
      "epoch": 12.35,
      "learning_rate": 1.91358024691358e-05,
      "loss": 2.099,
      "step": 1000
    },
    {
      "epoch": 18.52,
      "learning_rate": 3.7037037037037037e-06,
      "loss": 1.8147,
      "step": 1500
    }
  ],
  "logging_steps": 500,
  "max_steps": 1620,
  "num_train_epochs": 20,
  "save_steps": 500,
  "total_flos": 1553642422272000.0,
  "trial_name": null,
  "trial_params": null
}