OpenSERP-v1.py

from datasets import (
    DatasetBuilder,
    GeneratorBasedBuilder,
    DatasetInfo,
    Sequence,
    Split,
    SplitGenerator,
    DownloadManager,
    Features,
    Value,
    Version,
)
import os


_DESCRIPTION = """\
OpenSERP-v1 is a large dataset of over one billion embeddings designed to power a webscale information retrieval systems.
"""


class MyDataset(GeneratorBasedBuilder):
    VERSION = Version("1.0.0")

    def _info(self):
        return DatasetInfo(
            description=_DESCRIPTION,
            # Specify dataset features here
            features=Features(
                {
                    "id": Value("string"),
                    "url": Value("string"),
                    "title": Value("string"),
                    "metadata": Value("string"),
                    "dataset": Value("string"),
                    "text_chunks": Sequence(Value("string")),
                    "embeddings": Sequence(Value("float")),
                }
            ),
        )

    def _split_generators(self, dl_manager: DownloadManager):
        # Define dataset splits and where to find them
        return [
            SplitGenerator(
                name=Split.TRAIN,
                gen_kwargs={"filepath": "arxiv"},
            ),
            # Add other splits/folders as needed
        ]

    def _generate_examples(self, filepath):
        # Generate data examples from files
        for filename in os.listdir(filepath):
            file_path = os.path.join(filepath, filename)
            with open(file_path, encoding="utf-8") as f:
                data = f.read()
                # Process and yield each example
                yield filename, {"text": data}