trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.8666666666666667,
  "eval_steps": 50,
  "global_step": 14,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.33,
      "learning_rate": 5e-05,
      "loss": 1.5403,
      "step": 10
    },
    {
      "epoch": 1.87,
      "step": 14,
      "total_flos": 1.6333007128363008e+16,
      "train_loss": 1.4752462591443742,
      "train_runtime": 68.8091,
      "train_samples_per_second": 13.545,
      "train_steps_per_second": 0.203
    }
  ],
  "logging_steps": 10,
  "max_steps": 14,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 2,
  "save_steps": 100,
  "total_flos": 1.6333007128363008e+16,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}