generate_reference_files.py

import os
from glob import glob
from datasets import load_dataset

os.makedirs('./reference_files', exist_ok=True)


for split in ['validation', 'test']:
    dataset = load_dataset('asahi417/qg_ruquad', split=split, download_mode='force_redownload')
    for data in ['question', 'answer', 'sentence', 'paragraph']:
        with open('./reference_files/{}-{}.txt'.format(data, split), 'w') as f:
            if data == 'paragraph':
                tmp_data = dataset['paragraph_id']
            else:
                tmp_data = dataset[data]
            f.write('\n'.join([i.replace('\n', '.') for i in tmp_data]))

    length = [len(open(i).read().split('\n')) for i in glob(f'reference_files/*{split}.txt')]
    assert len(list(set(length))) == 1, length
    assert length[0] == len(dataset), f"{length[0]} != {len(dataset)}"