trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 2814,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.53,
      "learning_rate": 4.111584932480455e-05,
      "loss": 3.3646,
      "step": 500
    },
    {
      "epoch": 1.07,
      "learning_rate": 3.2231698649609096e-05,
      "loss": 0.932,
      "step": 1000
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.3347547974413646e-05,
      "loss": 0.3383,
      "step": 1500
    },
    {
      "epoch": 2.13,
      "learning_rate": 1.4463397299218195e-05,
      "loss": 0.1763,
      "step": 2000
    },
    {
      "epoch": 2.67,
      "learning_rate": 5.579246624022743e-06,
      "loss": 0.0893,
      "step": 2500
    },
    {
      "epoch": 3.0,
      "step": 2814,
      "total_flos": 1494191577600000.0,
      "train_loss": 0.8795227316481558,
      "train_runtime": 626.0855,
      "train_samples_per_second": 71.875,
      "train_steps_per_second": 4.495
    }
  ],
  "logging_steps": 500,
  "max_steps": 2814,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 1494191577600000.0,
  "trial_name": null,
  "trial_params": null
}