trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "global_step": 2193,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.68,
      "learning_rate": 3.860009119927041e-05,
      "loss": 2.3341,
      "step": 500
    },
    {
      "epoch": 1.37,
      "learning_rate": 2.7200182398540814e-05,
      "loss": 1.9037,
      "step": 1000
    },
    {
      "epoch": 2.05,
      "learning_rate": 1.580027359781122e-05,
      "loss": 1.8389,
      "step": 1500
    },
    {
      "epoch": 2.74,
      "learning_rate": 4.400364797081624e-06,
      "loss": 1.7612,
      "step": 2000
    },
    {
      "epoch": 3.0,
      "step": 2193,
      "train_runtime": 1890.7444,
      "train_samples_per_second": 1.16
    }
  ],
  "max_steps": 2193,
  "num_train_epochs": 3,
  "total_flos": 4319767004737536,
  "trial_name": null,
  "trial_params": null
}