trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "eval_steps": 500,
  "global_step": 1616,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.62,
      "learning_rate": 6.905940594059406e-05,
      "loss": 1.8011,
      "step": 500
    },
    {
      "epoch": 1.24,
      "learning_rate": 3.811881188118812e-05,
      "loss": 1.4847,
      "step": 1000
    },
    {
      "epoch": 1.86,
      "learning_rate": 7.178217821782178e-06,
      "loss": 1.4343,
      "step": 1500
    },
    {
      "epoch": 2.0,
      "step": 1616,
      "total_flos": 1.0496568754962432e+16,
      "train_loss": 1.563031800902716,
      "train_runtime": 2354.5427,
      "train_samples_per_second": 43.919,
      "train_steps_per_second": 0.686
    }
  ],
  "logging_steps": 500,
  "max_steps": 1616,
  "num_train_epochs": 2,
  "save_steps": 500,
  "total_flos": 1.0496568754962432e+16,
  "trial_name": null,
  "trial_params": null
}