trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 1437,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.04,
      "learning_rate": 3.26026443980515e-05,
      "loss": 1.3513,
      "step": 500
    },
    {
      "epoch": 2.09,
      "learning_rate": 1.5205288796102993e-05,
      "loss": 0.8309,
      "step": 1000
    },
    {
      "epoch": 3.0,
      "step": 1437,
      "total_flos": 2001462101114880.0,
      "train_loss": 0.9368828313584616,
      "train_runtime": 407.1166,
      "train_samples_per_second": 35.297,
      "train_steps_per_second": 3.53
    }
  ],
  "max_steps": 1437,
  "num_train_epochs": 3,
  "total_flos": 2001462101114880.0,
  "trial_name": null,
  "trial_params": null
}