download_miracl_corpus.py

import os
from huggingface_hub import hf_hub_download

# ALL langauges
langs = {
    'train': ['ar', 'bn', 'en', 'es', 'fa', 'fi', 'fr', 'hi', 'id', 'ja', 'ko', 'ru', 'sw', 'te', 'th', 'zh'],
    'dev': ['ar', 'bn', 'en', 'es', 'fa', 'fi', 'fr', 'hi', 'id', 'ja', 'ko', 'ru', 'sw', 'te', 'th', 'zh'],
    'testA': ['ar', 'bn', 'en', 'fi', 'id', 'ja', 'ko', 'ru', 'sw', 'te', 'th']
}

# the langauges we used in ESSIR
# the miracl corpus has been spitted into shards
languages2filesize = {'en': 66 , 'fa': 5, 'ru': 20, 'zh': 10}

for lang in languages2filesize.keys():
    for split in range(languages2filesize[lang]):
        hf_hub_download('miracl/miracl-corpus', 
            filename=f'docs-{split}.jsonl.gz', 
            subfolder=f'miracl-corpus-v1.0-{lang}', 
            repo_type='dataset', 
            cache_dir='/home/dju/datasets/essir-xlir/miracl/archived',
            force_filename=f'miracl-{lang}-{split}.jsonl.gz'
        )
        cmd = f'gunzip /home/dju/datasets/essir-xlir/miracl/archived/*gz'
        os.system(cmd)

cmd = f'rm -rvf /home/dju/datasets/essir-xlir/miracl/archived/*lock'
os.system(cmd)

for lang in languages2filesize.keys():
    # merge into a big file
    cmd = f'cat /home/dju/datasets/essir-xlir/miracl/archived/miracl-{lang}*.jsonl > /home/dju/datasets/essir-xlir/miracl/miracl-{lang}.jsonl'
    os.system(cmd)

    # remove shards
    cmd = f'rm -rf /home/dju/datasets/essir-xlir/miracl/archived/miracl-{lang}*.jsonl'