raw_redpajamas.py

"""Raw RedPajamas Custom Loading Script"""

import json
import datasets
from huggingface_hub import HfFileSystem


_CITATION = """"""
_DESCRIPTION = """"""
_HOMEPAGE = ""
_LICENSE = ""


class RawRedPajamas(datasets.GeneratorBasedBuilder):
    VERSION = datasets.Version("1.1.0")

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name="contents_only", version=VERSION, description="This returns the content field only."
        ),
    ]

    DEFAULT_CONFIG_NAME = "contents_only"  # It's not mandatory to have a default configuration. Just use one if it make sense.

    def _info(self):
        features = datasets.Features({"content": datasets.Value("string")})

        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
            homepage=_HOMEPAGE,
            license=_LICENSE,
            citation=_CITATION,
        )

    def _split_generators(self, dl_manager):
        fs = HfFileSystem(token=self.token)
        files = [file.split(f"/{self.repo_id}/")[-1] for file in fs.glob(f"datasets/{self.repo_id}/data/*.json")]
        downloaded_files = dl_manager.download_and_extract(files)
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                # These kwargs will be passed to _generate_examples
                gen_kwargs={
                    "filepaths": downloaded_files,
                },
            ),
        ]

    # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
    def _generate_examples(self, filepaths):
        for filepath in filepaths:
            with open(filepath, encoding="utf-8") as f:
                for key, row in enumerate(f):
                    data = json.loads(row)

                    content = data["raw_content"]

                    yield key, {
                        "content": content.strip().rstrip().lstrip(),
                    }