methods/CruxEval/parallel_executor.py

import os
import json
from datetime import datetime
from multiprocessing import Pool, cpu_count
from tqdm import tqdm
import logging
from typing import List, Optional
import random
import re

def is_decimal(s):
    try:
        a = float(s)
    except:
        return False
        
    return bool(re.match(r"^-?\d+\.\d+$", s))

def get_testcases(testcase_path):
    data = []
    if not os.path.exists(testcase_path):
        return []
    with open(testcase_path, 'r', encoding='utf-8') as f:
        for line in f:
            if line.strip():
                data.append(json.loads(line))
    return data

def check_difference(arr):
    for num in arr:
        if not is_decimal(num):
            return False
    # 遍历每一对元素
    for i in range(len(arr)):
        for j in range(i + 1, len(arr)):
            # 计算差异并判断是否满足条件
            if abs(float(arr[i]) - float(arr[j])) > 1e-6:
                return False  # 如果有任何一对差异大于 1e-6，则返回 False
    return True  # 所有对的差异都不大于 1e-6，返回 True

def append_dict_to_jsonl(file_path, data_dict):
    with open(file_path, 'a', encoding='utf-8') as f:
        f.write(json.dumps(data_dict, ensure_ascii=False) + '\n')


def setup_logging():
    os.makedirs("logs", exist_ok=True)
    log_file = f"logs/execution_{datetime.now().strftime('%Y%m%d_%H%M%S')}.log"
    
    logging.basicConfig(
        level=logging.INFO,
        format='%(asctime)s - %(levelname)s - %(message)s',
        handlers=[
            logging.FileHandler(log_file),
            logging.StreamHandler()
        ]
    )
    return logging.getLogger()    


logger = logging.getLogger(__name__)

from execute_tool import run_cpp_code_linux
def process_code_with_logging(data_item):
    """包装函数，用于添加日志"""
    problem_id = data_item["problem_id"]
    try:
        result = run_cpp_code_linux(data_item)
        status = result.get("error", "Unknown")
        logger.info(f"执行完成 - 问题ID: {problem_id}, 状态: {status}")
        return result
    except Exception as e:
        logger.error(f"执行异常 - 问题ID: {problem_id}, 错误: {str(e)}")
        data_item["error"] = ["EXE"]
        data_item["details"] = str(e)
        return data_item


from load_response import get_response_function, load_data, load_qwen3_result
if __name__ == "__main__":
    datasets_name = "tcb"
    testcase_alg = "crux"
    model_name = "Qwen2.5-32B-Instruct"

    if not os.path.exists(f"/home/luoxianzhen/yang/save_tests_{model_name}"):
        os.mkdir(f"/home/luoxianzhen/yang/save_tests_{model_name}")
    if not os.path.exists(f"/home/luoxianzhen/yang/save_tests_{model_name}/{testcase_alg}/"):
        os.mkdir(f"/home/luoxianzhen/yang/save_tests_{model_name}/{testcase_alg}/")
    
    test_dir = f"/home/luoxianzhen/yang/save_tests_{model_name}/{testcase_alg}/" + "tests-{}.jsonl"
    pass_rate_save_file = f"/home/luoxianzhen/yang/save_tests_{model_name}/{testcase_alg}/test_pass_rate.jsonl"

    logger = setup_logging()
    logger.info("开始执行代码评估...")

    cpu = 50
    # cpu = 2

    logger.info(f" === Stage 1 == 整理输出")
    import time

    data = get_response_function(repsonse_path=f"/home/luoxianzhen/yang/data/response-orginal/orginal_response_{testcase_alg}_{model_name}.jsonl", model_name=model_name)

    logger.info(f"使用 {cpu} 个CPU核心进行并行处理 === Stage 2 == 执行得到输出")

    testcase_data = load_data(data)

    with Pool(cpu) as pool:
        results = list(tqdm(
            pool.imap_unordered(process_code_with_logging, testcase_data),
            total=len(testcase_data),
            desc="执行进度"
        ))
    logger.info("所有代码执行完成 - 开始保存筛选后的测试样例")

    result_dict = {}
    for item in results:
        if item['problem_id'] not in result_dict.keys():
            result_dict[item['problem_id']] = [item, ]
        else:
            result_dict[item['problem_id']].append(item)

    for k, v in result_dict.items():
        testcases = data[k]
        save_path = test_dir.format(k)
        status_array_length = len(v[0]["error"])
        saved_nums = 0
        for i in range(status_array_length):
            all_AC = True 
            output_list = []
            for item in v:
                if i >= len(item["error"]):
                    continue
                if item["error"][i] == "EXE":
                    all_AC = False
                    break

                output_list.append(item['details'][i])
                if not item["error"][i] == "AC":
                    all_AC = False

            if all_AC and (len(list(set(output_list))) == 1 or check_difference(output_list)):
                if len(output_list) <= 0 or output_list[0] == "":
                    continue
                append_dict_to_jsonl(save_path, testcases[i])
                saved_nums += 1

        append_dict_to_jsonl(pass_rate_save_file, {
            "tcb_id": k,
            "gen_nums": len(testcases),
            "right_nums": saved_nums
        })