对比不同 LLM 模型的 token 消耗
(
date: str,
papers: List[Dict],
user_profiles: List[Dict],
models: List[str] = None,
)
| 336 | |
| 337 | def compare_llm_models( |
| 338 | date: str, |
| 339 | papers: List[Dict], |
| 340 | user_profiles: List[Dict], |
| 341 | models: List[str] = None, |
| 342 | ) -> List[Dict[str, Any]]: |
| 343 | """对比不同 LLM 模型的 token 消耗""" |
| 344 | if models is None: |
| 345 | models = ["qwen3.5-plus", "qwen-plus", "qwen-max"] |
| 346 | |
| 347 | results = [] |
| 348 | for model in models: |
| 349 | print(f"Benchmarking {model}...") |
| 350 | result = benchmark_one_day(date, model, papers, user_profiles) |
| 351 | results.append(result) |
| 352 | |
| 353 | # 清空日志以便下次测试 |
| 354 | if TOKEN_LOG_PATH.exists(): |
| 355 | TOKEN_LOG_PATH.unlink() |
| 356 | |
| 357 | return results |
| 358 | |
| 359 | |
| 360 | def main(): |
| 361 | parser = argparse.ArgumentParser(description="Token 花销统计工具") |
no test coverage detected