trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "global_step": 36,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.11,
      "learning_rate": 0.00014444444444444444,
      "loss": 0.182,
      "step": 10
    },
    {
      "epoch": 2.22,
      "learning_rate": 8.888888888888889e-05,
      "loss": 0.0101,
      "step": 20
    },
    {
      "epoch": 3.33,
      "learning_rate": 3.3333333333333335e-05,
      "loss": 0.0058,
      "step": 30
    },
    {
      "epoch": 4.0,
      "step": 36,
      "total_flos": 4.463538601800499e+16,
      "train_loss": 0.05584914175172647,
      "train_runtime": 136.48,
      "train_samples_per_second": 4.22,
      "train_steps_per_second": 0.264
    }
  ],
  "max_steps": 36,
  "num_train_epochs": 4,
  "total_flos": 4.463538601800499e+16,
  "trial_name": null,
  "trial_params": null
}