import json
list = []
paragraph = []
with open('./icliniq-21k.json', mode='r', encoding='utf-8') as f:
json_list = json.load(f)
for index, dict in enumerate(json_list):
quary = 'instruction: ' + dict['instruction'] + ' input: ' + dict['input']
answer = ['output: ' + dict['output']] # a
paragraph.append({"q": quary, "a": answer}) # 一条对话信息添加到paragraph
qa = {"id": 0, "paragraph":paragraph}
list.append(qa)
paragraph = []
with open('./Sicliniq_21k.json',mode='w',encoding='utf-8',newline='\n') as f:
index = 0
for i in list:
# print(i)
index += 1
i['id'] = index
f.write(json.dumps(i, ensure_ascii=False) + '\n')
手写python脚本转换格式(不推荐),推荐使用alpaca2qa.py套模板转换格式.