luna-speech-dataset_depr.py

# coding=utf-8
"""LUNA Speech Dataset"""

import pathlib

import datasets


_DESCRIPTION = """\
TODO
"""
_HOMEPAGE = ""
_CITATION = ""
_LICENSE = "2-Clause BSD"



class LUNA(datasets.GeneratorBasedBuilder):

    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(
                {
                    "audio": datasets.Audio(sampling_rate=16_000),
                    "transcript": datasets.Value("string"),
                    "gender": datasets.Value("string"),
                    "id": datasets.Value("string"),
                },
            ),
            supervised_keys=None,
            homepage=_HOMEPAGE,
            citation=_CITATION,
            license=_LICENSE,
        )

    def _split_generators(self, dl_manager):
        metadata_path = dl_manager.download_and_extract(
            "https://huggingface.co./datasets/czyzi0/luna-speech-dataset/raw/main/metadata.tsv"
        )
        wavs_path = dl_manager.download_and_extract(
            "https://huggingface.co./datasets/czyzi0/luna-speech-dataset/resolve/main/wavs.tar.gz"
        )
        return [
            datasets.SplitGenerator(
                name="train",
                gen_kwargs={"metadata_path": metadata_path, "wavs_path": wavs_path}
            )
        ]

    def _generate_examples(self, metadata_path, wavs_path):
        wavs_path = pathlib.Path(wavs_path)
        with open(metadata_path, "r") as fh:
            header = next(fh).strip().split("\t")
            for item_idx, line in enumerate(fh):
                line = line.strip().split("\t")
                id_ = line[header.index("id")]
                transcript = line[header.index("transcript")]
                gender = line[header.index("gender")]

                wav_path = wavs_path / "wavs" / f"{id_}.wav"
                with open(wav_path, "rb") as fh_:
                    item = {
                        "audio": {"path": str(wav_path.absolute()), "bytes": fh_.read()},
                        "transcript": transcript,
                        "gender": gender,
                        "id": id_,
                    }
                    yield item_idx, item