lb-data-test.py

from typing import Tuple
from typing import Generator

import datasets
from datasets import Value

_DESCRIPTION = """\
...
"""

_URL = "..."


class LaboASRConfig(datasets.BuilderConfig):

    def __init__(
        self,
        do_stuff: bool = False,
        **kwargs,
    ):
        super(LaboASRConfig, self).__init__(version=datasets.Version("0.0.2", ""), **kwargs)
        self.do_stuff = do_stuff


class LaboASR(datasets.GeneratorBasedBuilder):

    CORTI_DATASET_NAME = "labo"
    DEFAULT_WRITER_BATCH_SIZE = 256
    DEFAULT_CONFIG_NAME = "default"
    BUILDER_CONFIG_CLASS = LaboASRConfig

    BUILDER_CONFIGS = [
        LaboASRConfig(
            name="default",
            description="Default config.",
        ),
        LaboASRConfig(
            name="alternative",
            description="Alternative config.",
        ),
    ]

    def _info(self):
        features = {
            "transcript": Value("string"),
            "id": Value("string"),
        }
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=datasets.Features(features),
            homepage=_URL,
        )

    def _split_generators(self, dl_manager):
        
        if self.config.name == "default":
            return [
                datasets.SplitGenerator(
                    name=datasets.Split.TRAIN,
                    gen_kwargs={
                        "paths": ["1", "2", "3", "8", "9", "10"],
                    },
                ),
                datasets.SplitGenerator(
                    name=datasets.Split.VALIDATION,
                    gen_kwargs={
                        "paths": ["4", "5", "6", "7"],
                    },
                ),
            ]
        elif self.config.name == "alternative":
            return [
                datasets.SplitGenerator(
                    name=datasets.Split.TRAIN,
                    gen_kwargs={
                        "paths": ["1", "2", "3", "4"],
                    },
                ),
                datasets.SplitGenerator(
                    name=datasets.Split.VALIDATION,
                    gen_kwargs={
                        "paths": ["5", "6", "7"],
                    },
                ),
            ]

    def _generate_examples(
        self,
        paths: list[str],
    ) -> Generator[Tuple[int, dict], None, None]:

        for i, p in enumerate(paths):

            # Yield example
            example = {
                "transcript": "Hello, world!",
                "id": "0000" + p,
            }

            yield i, example