Source code for torchtext.datasets.sogounews

import os
from typing import Union, Tuple

from torchtext._internal.module_utils import is_module_available
from torchtext.data.datasets_utils import (
    _wrap_split_argument,
    _create_dataset_directory,
)

if is_module_available("torchdata"):
    from torchdata.datapipes.iter import FileOpener, GDriveReader, IterableWrapper


URL = "https://drive.google.com/uc?export=download&id=0Bz8a_Dbh9QhbUkVqNEszd0pHaFE"

MD5 = "0c1700ba70b73f964dd8de569d3fd03e"

NUM_LINES = {
    "train": 450000,
    "test": 60000,
}

_PATH = "sogou_news_csv.tar.gz"

_EXTRACTED_FILES = {
    "train": os.path.join("sogou_news_csv", "train.csv"),
    "test": os.path.join("sogou_news_csv", "test.csv"),
}

_EXTRACTED_FILES_MD5 = {
    "train": "f36156164e6eac2feda0e30ad857eef0",
    "test": "59e493c41cee050329446d8c45615b38",
}

DATASET_NAME = "SogouNews"


@_create_dataset_directory(dataset_name=DATASET_NAME)
@_wrap_split_argument(("train", "test"))
def SogouNews(root: str, split: Union[Tuple[str], str]):
    """SogouNews Dataset

   For additional details refer to https://arxiv.org/abs/1509.01626

    Number of lines per split:
        - train: 450000
        - test: 60000

    Args:
        root: Directory where the datasets are saved. Default: os.path.expanduser('~/.torchtext/cache')
        split: split or splits to be returned. Can be a string or tuple of strings. Default: (`train`, `test`)

    :returns: DataPipe that yields tuple of label (1 to 5) and text containing the news title and contents
    :rtype: (int, str)
    """
    if not is_module_available("torchdata"):
        raise ModuleNotFoundError(
            "Package `torchdata` not found. Please install following instructions at `https://github.com/pytorch/data`"
        )

    url_dp = IterableWrapper([URL])
    cache_compressed_dp = url_dp.on_disk_cache(
        filepath_fn=lambda x: os.path.join(root, _PATH),
        hash_dict={os.path.join(root, _PATH): MD5},
        hash_type="md5",
    )
    cache_compressed_dp = GDriveReader(cache_compressed_dp).end_caching(
        mode="wb", same_filepath_fn=True
    )

    cache_decompressed_dp = cache_compressed_dp.on_disk_cache(
        filepath_fn=lambda x: os.path.join(root, _EXTRACTED_FILES[split])
    )
    cache_decompressed_dp = (
        FileOpener(cache_decompressed_dp, mode="b")
        .read_from_tar()
        .filter(lambda x: _EXTRACTED_FILES[split] in x[0])
    )
    cache_decompressed_dp = cache_decompressed_dp.end_caching(
        mode="wb", same_filepath_fn=True
    )

    data_dp = FileOpener(cache_decompressed_dp, mode="b")
    return data_dp.parse_csv().map(fn=lambda t: (int(t[0]), " ".join(t[1:])))