Function main

examples/run_benchmark.py:361–393

()

async def main():
    parser = argparse.ArgumentParser(description='Test Benchmark Loop')
    parser.add_argument("--config", default=os.path.join(root, "configs", "tool_calling_agent.py"), help="config file path")
    parser.add_argument("--benchmark", default="leetcode", help="benchmark name to test")
    parser.add_argument("--max-concurrency", type=int, default=4, help="Maximum number of concurrent tasks (default: 5)")
    parser.add_argument(
        '--cfg-options',
        nargs='+',
        action=DictAction,
        help='override settings')
    args = parser.parse_args()
    
    config.initialize(config_path=args.config, args=args)
    logger.initialize(config=config)
    
    logger.info("| 🧠 Initializing model manager...")
    if hasattr(model_manager, 'initialize'):
        await model_manager.initialize()
    
    benchmark_name = args.benchmark
    logger.info(f"| 🛠️ Initializing benchmark manager for {benchmark_name}...")
    await benchmark_manager.initialize(benchmark_names=[benchmark_name])

    # Initialize result saver
    logger.info(f"| 💾 Initializing result saver...")
    result_saver = BenchmarkResultSaver(benchmark_name, args.max_concurrency, 0, TARGET_MODEL)  # We'll update total_tasks later
    logger.info(f"| ✅ Results will be saved to: {result_saver.get_file_path()}")

    await test_math_benchmark(benchmark_name, max_concurrency=args.max_concurrency, result_saver=result_saver)
    
    print("| 🧹 Cleaning up...")
    await benchmark_manager.cleanup()
    print("| 🚪 Test completed")

if __name__ == "__main__":
    asyncio.run(main())