checkpoint-1000/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.21482277121374865,
  "global_step": 1000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.11,
      "learning_rate": 6.10123648e-05,
      "loss": 1.2324,
      "step": 500
    },
    {
      "epoch": 0.21,
      "learning_rate": 5.472524835555556e-05,
      "loss": 0.7401,
      "step": 1000
    }
  ],
  "max_steps": 5000,
  "num_train_epochs": 2,
  "total_flos": 4.61736640512e+18,
  "trial_name": null,
  "trial_params": null
}