trainer_state.json

{
  "best_global_step": 44,
  "best_metric": 1.7750476598739624,
  "best_model_checkpoint": "./mcqa_qwen3_letter_only_preference/checkpoint-44",
  "epoch": 1.0,
  "eval_steps": 500,
  "global_step": 44,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "eval_loss": 1.7750476598739624,
      "eval_runtime": 3.0077,
      "eval_samples_per_second": 43.554,
      "eval_steps_per_second": 5.652,
      "step": 44
    }
  ],
  "logging_steps": 50,
  "max_steps": 88,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 2,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 332517691883520.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}