add_instruction_to_single_file.py

import os
import pandas as pd
import json
from tqdm import tqdm
from glob import glob
from collections import defaultdict

# root_path = '/etc/ssd1/jiangzhongtao/baai_embedding_tune/data/all_collect/mteb-classification'
# fnames = os.listdir(root_path)
# dataset2size = defaultdict(int)
# for fname in fnames:
#     _, name, lang = fname.split('_')
#     fname = os.path.join(root_path, fname)
#     with open(fname, 'r') as f:
#         for line in f:
#             dataset2size[name] += 1

src = 'mteb-Clustering'
# postfix = '_hn'

with open(f'/etc/ssd1/jiangzhongtao/baai_embedding_tune/data/all_collect/instructions/{src}_instructions.json') as f:
    instructions = json.load(f)

# src += postfix
root_path = f'/etc/ssd1/jiangzhongtao/baai_embedding_tune/data/all_collect/{src}'
fnames = os.listdir(root_path)
pbar = tqdm()
for fname in tqdm(fnames):
    _, name, lang = fname[:-6].split('_')
    instruction = instructions[name]
    f_in = open(os.path.join(root_path, fname), 'r')
    f_out = open(os.path.join(root_path, fname) + '.inst', 'w')
    for line in f_in:
        line = json.loads(line)
        line['instruction'] = instruction
        line = json.dumps(line, ensure_ascii=False)
        f_out.write(line + '\n')
        pbar.update(1)
    f_in.close()
    f_out.close()

for fname in tqdm(glob(f'/etc/ssd1/jiangzhongtao/baai_embedding_tune/data/all_collect/{src}/*.jsonl')):
    os.remove(fname)

for fname in tqdm(glob(f'/etc/ssd1/jiangzhongtao/baai_embedding_tune/data/all_collect/{src}/*.jsonl.inst')):
    os.rename(fname, fname.replace('.inst', ''))