trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 234,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.64,
      "learning_rate": 1.5726495726495726e-05,
      "loss": 0.6931,
      "step": 50
    },
    {
      "epoch": 1.28,
      "learning_rate": 1.1452991452991454e-05,
      "loss": 0.6785,
      "step": 100
    },
    {
      "epoch": 1.92,
      "learning_rate": 7.17948717948718e-06,
      "loss": 0.6608,
      "step": 150
    },
    {
      "epoch": 2.56,
      "learning_rate": 2.9059829059829063e-06,
      "loss": 0.5927,
      "step": 200
    },
    {
      "epoch": 3.0,
      "step": 234,
      "total_flos": 982737415710720.0,
      "train_loss": 0.6440325027857071,
      "train_runtime": 111.2546,
      "train_samples_per_second": 67.143,
      "train_steps_per_second": 2.103
    }
  ],
  "logging_steps": 50,
  "max_steps": 234,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 982737415710720.0,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}