SEA_mC4.py

"""South East Asia mC4 dataset."""
import gzip
import json
import datasets

logger = datasets.logging.get_logger(__name__)
_DESCRIPTION = """
South East Asia mC4 dataset."""
_CITATION = """EMPTY"""
_URL = "EMPTY"
_DATA_URL = "https://huggingface.co/datasets/aisingapore/sea-pile/tree/main/sea-pile-mc4/{language}/mc4-{language}-{index:05d}-of-{n_shards:05d}.json.gz"

_N_SHARDS_PER_LANGUAGES = {
    "zh": 468,
    "id": 21,
    "ms": 4,
    "tl": 6,
    "my": 11,
    "vi": 329,
    "th": 74,
    "lo": 2,
    "km": 9,
    "ta": 29,
}


class SEAPileConfig(datasets.BuilderConfig):
    """BuilderConfig for SEAmC4."""

    def __init__(self, *args, languages, **kwargs):
        """BuilderConfig for SEAmC4.
        Args:
            languages (:obj:`List[str]`): list of languages to load
            **kwargs: keyword arguments forwarded to super.
        """
        super().__init__(
            *args,
            name="+".join(languages),
            **kwargs,
        )
        self.languages = languages


class SEAPile(datasets.GeneratorBasedBuilder):
    """South East Asia mC4 dataset."""

    BUILDER_CONFIGS = [
        SEAPileConfig(languages=[lang]) for lang in _N_SHARDS_PER_LANGUAGES
    ]
    BUILDER_CONFIG_CLASS = SEAPileConfig

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "id": datasets.Value("string"),
                    "text": datasets.Value("string"),
                }
            ),
            supervised_keys=None,
            homepage=_URL,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        data_urls = {}
        for split in ["train"]:
            data_urls[split] = [
                _DATA_URL.format(
                    language=lang,
                    index=index,
                    n_shards=_N_SHARDS_PER_LANGUAGES[lang],
                )
                for lang in self.config.languages
                for index in range(0, _N_SHARDS_PER_LANGUAGES[lang])
            ]
        train_downloaded_files = dl_manager.download(data_urls["train"])

        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                gen_kwargs={"filepaths": train_downloaded_files},
            ),
        ]

    def _generate_examples(self, filepaths):
        """This function returns the examples in the raw (text) form by iterating on all the files."""
        id_ = 0
        for filepath in filepaths:
            logger.info("generating examples from = %s", filepath)
            with gzip.open(open(filepath, "rb"), "rt", encoding="utf-8") as f:
                for line in f:
                    if line:
                        example = json.loads(line)
                        yield id_, example
                        id_ += 1