mimir-core.py

"""Mímir Core v1 dataset."""
import gzip
import json
import datasets

logger = datasets.logging.get_logger(__name__)
_DESCRIPTION = """\\nMímir Core v1."""
_CITATION = """
"""
_URL = "https://github.com/NbAiLab/mimir-data"
_DATA_URL = "https://huggingface.co/datasets/mimir-project/mimir-core/resolve/main/data/{split_suffix}-{segment}-{index:04d}-of-{n_shards:04d}.json"
_N_SHARDS_PER_SPLIT = {
    "bad": {"train": 6, "validation": 1},
    "medium": {"train": 21, "validation": 1},
    "good": {"train": 7, "validation": 1},
}
_SEGMENTS = ("bad", "medium", "good")


class MimirCoreConfig(datasets.BuilderConfig):
    """BuilderConfig for MimirCore."""

    def __init__(self, name=None, *args, **kwargs):
        """BuilderConfig for MimirCore.
        Args:
            **kwargs: keyword arguments forwarded to super.
        """
        if name is None:
            name = "default"
        elif name not in _SEGMENTS:
            raise ValueError(f"Invalid segment option '{name}'. Options are {str(_SEGMENTS)}.")
        self.name = name
        super().__init__(
            *args,
            name=name,
            **kwargs,
        )


class MimirCore(datasets.GeneratorBasedBuilder):
    """Mimir Core v1."""
    BUILDER_CONFIGS = [MimirCoreConfig()] + [MimirCoreConfig(segment) for segment in _SEGMENTS]
    BUILDER_CONFIG_CLASS = MimirCoreConfig
    DEFAULT_CONFIG_NAME = "default"

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "id": datasets.Value("string"),
                    "doc_type": datasets.Value("string"),
                    "publish_year":  datasets.Value("int32"),
                    "lang_fasttext":  datasets.Value("string"),
                    "lang_fasttext_conf":  datasets.Value("string"),
                    "text":  datasets.Value("string"),
                    "perplexity": datasets.Value("float"),
                    "perplexity_model": datasets.Value("string"),
                    "harmful_pp": datasets.Value("float"),
                    "segment": datasets.Value("string"),
                }
            ),
            supervised_keys=None,
            homepage=_URL,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        if self.config.name != "default":
            segments = [self.config.name]
        else:
            segments = _SEGMENTS
        data_urls = {}
        for split in ["train", "validation"]:
            data_urls[split] = []
            for segment in segments:
                data_urls[split] += [
                    _DATA_URL.format(
                        split_suffix=split,
                        segment=segment,
                        index=index,
                        n_shards=_N_SHARDS_PER_SPLIT[segment][split],
                    )
                    for index in range(1, _N_SHARDS_PER_SPLIT[segment][split] + 1)
                ]
        train_downloaded_files = dl_manager.download(data_urls["train"])
        validation_downloaded_files = dl_manager.download(data_urls["validation"])

        return [
            datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={"filepaths": train_downloaded_files}),
            datasets.SplitGenerator(
                name=datasets.Split.VALIDATION, gen_kwargs={"filepaths": validation_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 open(filepath, "rb") as b, gzip.open(b, "rt", encoding="utf-8") as f:
                for line in f:
                    if line.strip():
                        example = json.loads(line)
                        yield id_, example
                        id_ += 1