nbnn_language_detection.py

from datasets import DatasetBuilder, GenerateMode, SplitGenerator, load_dataset
from datasets.features import Features, Sequence, Value

class MyDataset(DatasetBuilder):

    VERSION = "0.1.0"

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

    def _split_generators(self, dl_manager):
        files = {
            'train': 'train.jsonl',
            'dev': 'dev.jsonl',
            'test': 'test.jsonl'
        }

        return [
            SplitGenerator(name=split, gen_kwargs={'filepath': path})
            for split, path in files.items()
        ]

    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']
                }