trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "global_step": 2140,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.93,
      "learning_rate": 7.663551401869159e-06,
      "loss": 0.5021,
      "step": 500
    },
    {
      "epoch": 1.87,
      "learning_rate": 5.3271028037383174e-06,
      "loss": 0.3487,
      "step": 1000
    },
    {
      "epoch": 2.8,
      "learning_rate": 2.9906542056074766e-06,
      "loss": 0.2435,
      "step": 1500
    },
    {
      "epoch": 3.74,
      "learning_rate": 6.542056074766355e-07,
      "loss": 0.1797,
      "step": 2000
    },
    {
      "epoch": 4.0,
      "step": 2140,
      "total_flos": 204352668149760.0,
      "train_loss": 0.3096321917025842,
      "train_runtime": 1506.3037,
      "train_samples_per_second": 22.707,
      "train_steps_per_second": 1.421
    }
  ],
  "max_steps": 2140,
  "num_train_epochs": 4,
  "total_flos": 204352668149760.0,
  "trial_name": null,
  "trial_params": null
}