mteb-Retrieval_main-hn/test.py

src_files = '''
/etc/ssd1/dengjingcheng/general_embedding/mteb_all_data/retrieval/treccovid/mteb_TRECCOVID_en_hn.jsonl
/etc/ssd1/dengjingcheng/general_embedding/mteb_all_data/retrieval/fiqa/mteb_FiQA2018_en_hn.jsonl
/etc/ssd1/dengjingcheng/general_embedding/mteb_all_data/retrieval/nfcorpus/mteb_NFCorpus_en_hn.jsonl
/etc/ssd1/dengjingcheng/general_embedding/mteb_all_data/retrieval/scifact/mteb_SciFact_en_hn.jsonl
/etc/ssd1/dengjingcheng/general_embedding/mteb_all_data/retrieval/fever/mteb_FEVER_en_hn.jsonl
/etc/ssd1/dengjingcheng/general_embedding/mteb_all_data/retrieval/dbpedia/mteb_DBPedia_en_hn.jsonl
/etc/ssd1/dengjingcheng/general_embedding/mteb_all_data/retrieval/hotpotqa/mteb_HotpotQA_en_hn.jsonl
/etc/ssd1/dengjingcheng/general_embedding/mteb_all_data/retrieval/msmarco/mteb_MSMARCO_en_hn.jsonl
'''

src_files = src_files.strip().split('\n')

from tqdm import tqdm
import json
import os
import random

dst_path = '/etc/ssd1/jiangzhongtao/baai_embedding_tune/data/all_collect/mteb-Retrieval_hn'
for src_file in tqdm(src_files):
    _, fname = os.path.split(src_file)
    fname = fname.replace('_hn', '')
    dst_file = os.path.join(dst_path, fname)

    src_f = open(src_file, 'r')
    dst_f = open(dst_file, 'w')

    for line in src_f:
        line = json.loads(line)
        assert isinstance(line['neg'], list)
        if len(line['neg']) != 1:
            line['neg'] = random.sample(line['neg'][30:210], k=10)
            del line['sims']
        dst_f.write(json.dumps(line, ensure_ascii=False) + '\n')
    
    src_f.close()
    dst_f.close()