trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 2100,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.71,
      "learning_rate": 3.809523809523809e-05,
      "loss": 2.2496,
      "step": 500
    },
    {
      "epoch": 1.43,
      "learning_rate": 2.6190476190476192e-05,
      "loss": 1.6046,
      "step": 1000
    },
    {
      "epoch": 2.14,
      "learning_rate": 1.4285714285714285e-05,
      "loss": 1.275,
      "step": 1500
    },
    {
      "epoch": 2.86,
      "learning_rate": 2.3809523809523808e-06,
      "loss": 0.9903,
      "step": 2000
    },
    {
      "epoch": 3.0,
      "step": 2100,
      "total_flos": 3707289517522944.0,
      "train_loss": 1.5039557393391927,
      "train_runtime": 621.7573,
      "train_samples_per_second": 33.766,
      "train_steps_per_second": 3.378
    }
  ],
  "max_steps": 2100,
  "num_train_epochs": 3,
  "total_flos": 3707289517522944.0,
  "trial_name": null,
  "trial_params": null
}