src/download_culturax.py

import argparse
import json
from datasets import load_dataset

def download_culturax(language, output_file):
    # Load the CulturaX dataset for the specified language
    dataset = load_dataset("uonlp/CulturaX", language, use_auth_token=True)

    # Write a limited number of samples to the JSON Lines file
    with open(output_file, 'w', encoding='utf-8') as f:
        for example in dataset['train']:
            json.dump(example, f)
            f.write("\n")

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Download a part of CulturaX and save it as a JSON Lines file.")
    parser.add_argument('--language', type=str, required=True, help="The language code for the part of CulturaX to download (e.g., 'no' for Norwegian).")
    parser.add_argument('--output_file', type=str, required=True, help="The output JSON Lines file.")

    args = parser.parse_args()
    download_culturax(args.language, args.output_file)