nbnn_language_detection.py

from datasets import DatasetBuilder, DatasetInfo, SplitGenerator
from datasets.features import Features, Value
import json

class MyDataset(DatasetBuilder):

    VERSION = "0.1.0"

    def _info(self):
        return DatasetInfo(
            features=Features({
                'text': Value('string'),
                'language': Value('string'),
            }),
            supervised_keys=('text', 'language')
        )

    def _split_generators(self, dl_manager):
        urls = {
            'train': 'https://huggingface.co/datasets/NbAiLab/nbnn_language_detection/resolve/main/train.jsonl',
            'dev': 'https://huggingface.co/datasets/NbAiLab/nbnn_language_detection/resolve/main/dev.jsonl',
            'test': 'https://huggingface.co/datasets/NbAiLab/nbnn_language_detection/resolve/main/test.jsonl',
        }
        
        downloaded_files = dl_manager.download(urls)

        return [
            SplitGenerator(name=split, gen_kwargs={'filepath': downloaded_files[split]})
            for split in urls.keys()
        ]

    def _generate_examples(self, filepath):
        with open(filepath, 'r') as f:
            for id, line in enumerate(f):
                data = json.loads(line)
                yield id, {
                    'text': data['text'],
                    'language': data['language']
                }