data_format_prepare.py

from base_info import *
from tqdm import tqdm
import json


# 替换 response 中的中文
def replace_keys(d, mapping=cn_en_map):
    if isinstance(d, dict):
        # 替换字典中的键
        return {mapping.get(k, k): replace_keys(v, mapping) for k, v in d.items()}
    elif isinstance(d, list):
        # 如果是列表，递归处理每个元素
        return [mapping.get(item, item) for item in d]
    return d  # 处理其他类型



def format_prepare(input_path, output_path):
    data_all0 = []
    with open(input_path, 'r', encoding='utf-8') as file:
        for line in tqdm(file):
            data_json   = json.loads(line)
            data_all0.append(data_json)

    data_all = []
    for data_json0 in tqdm(data_all0):
        data_json = {
            "response": replace_keys(data_json0["response"], cn_en_map),
            "images": ["" + data_json0["images"]],
            "query": get_prompt(data_json0["prompt_ori"]),
            "data_source": data_json0["data_source"],
        }
        data_json["response"] = json.dumps(data_json["response"], ensure_ascii=False)
        data_all.append(data_json)
    
    #                     "images": [resized_img],
    #                     "problem": data["query"],
    #                     "answer": data["response"],

    # original_data["prompt_ori"] = replace_keys(original_data["prompt_ori"], cn_en_map)
    # original_data["prompt_ori"] = get_prompt(original_data["prompt_ori"])


    with open(output_path, 'w', encoding='utf-8') as file:
        for item in tqdm(data_all):
            file.write(json.dumps(item, ensure_ascii=False) + '\n')
    return




if __name__ == "__main__":
    input_path = ""
    output_path = ""

    # input_path = "./magic_mirror_data_test.jsonl"
    # output_path = "./magic_mirror_data_test_r1.jsonl"

    input_path = "./magic_mirror_data_train.jsonl"
    output_path = "./magic_mirror_data_train_r1.jsonl"

    format_prepare(input_path, output_path)