trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 2385,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.63,
      "learning_rate": 3.9517819706498955e-05,
      "loss": 0.3391,
      "step": 500
    },
    {
      "epoch": 1.26,
      "learning_rate": 2.9035639412997907e-05,
      "loss": 0.2391,
      "step": 1000
    },
    {
      "epoch": 1.89,
      "learning_rate": 1.8553459119496856e-05,
      "loss": 0.1777,
      "step": 1500
    },
    {
      "epoch": 2.52,
      "learning_rate": 8.071278825995808e-06,
      "loss": 0.1349,
      "step": 2000
    },
    {
      "epoch": 3.0,
      "step": 2385,
      "total_flos": 2578738654660608.0,
      "train_runtime": 2038.0134,
      "train_samples_per_second": 1.17
    }
  ],
  "max_steps": 2385,
  "num_train_epochs": 3,
  "total_flos": 2578738654660608.0,
  "trial_name": null,
  "trial_params": null
}