trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 1083,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.39,
      "learning_rate": 2.6915974145891044e-05,
      "loss": 3.4565,
      "step": 500
    },
    {
      "epoch": 2.77,
      "learning_rate": 3.831948291782087e-06,
      "loss": 2.8895,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "step": 1083,
      "total_flos": 4518261817344000.0,
      "train_loss": 3.1487963703750794,
      "train_runtime": 682.6239,
      "train_samples_per_second": 12.666,
      "train_steps_per_second": 1.587
    }
  ],
  "max_steps": 1083,
  "num_train_epochs": 3,
  "total_flos": 4518261817344000.0,
  "trial_name": null,
  "trial_params": null
}