Add support for MRPC dataset with unit tests (#1712)

* Add support for MRPC dataset
* Add unit tests
* Remove lambda functions
* Add dataset documentation
* Add shuffle and sharding

docs/source/datasets.rst

@@ -57,6 +57,11 @@ IMDb
 
 .. autofunction:: IMDB
 
+MRPC
+~~~~
+
+.. autofunction:: MRPC
+
 SogouNews
 ~~~~~~~~~
 

test/datasets/test_mrpc.py

@@ -0,0 +1,71 @@
+import os
+from collections import defaultdict
+from unittest.mock import patch
+
+from parameterized import parameterized
+from torchtext.datasets.mrpc import MRPC
+
+from ..common.case_utils import TempDirMixin, zip_equal, get_random_unicode
+from ..common.torchtext_test_case import TorchtextTestCase
+
+
+def _get_mock_dataset(root_dir):
+    """
+    root_dir: directory to the mocked dataset
+    """
+    base_dir = os.path.join(root_dir, "MRPC")
+    os.makedirs(base_dir, exist_ok=True)
+
+    seed = 1
+    mocked_data = defaultdict(list)
+    for file_name, file_type in [("msr_paraphrase_train.txt", "train"), ("msr_paraphrase_test.txt", "test")]:
+        txt_file = os.path.join(base_dir, file_name)
+        with open(txt_file, "w", encoding="utf-8") as f:
+            f.write("Quality\t#1 ID\t#2 ID\t#1 String\t#2 String\n")
+            for i in range(5):
+                label = seed % 2
+                rand_string_1 = get_random_unicode(seed)
+                rand_string_2 = get_random_unicode(seed + 1)
+                dataset_line = (label, rand_string_1, rand_string_2)
+                f.write(f"{label}\t{i}\t{i}\t{rand_string_1}\t{rand_string_2}\n")
+
+                # append line to correct dataset split
+                mocked_data[file_type].append(dataset_line)
+                seed += 1
+
+    return mocked_data
+
+
+class TestMRPC(TempDirMixin, TorchtextTestCase):
+    root_dir = None
+    samples = []
+
+    @classmethod
+    def setUpClass(cls):
+        super().setUpClass()
+        cls.root_dir = cls.get_base_temp_dir()
+        cls.samples = _get_mock_dataset(cls.root_dir)
+        cls.patcher = patch("torchdata.datapipes.iter.util.cacheholder._hash_check", return_value=True)
+        cls.patcher.start()
+
+    @classmethod
+    def tearDownClass(cls):
+        cls.patcher.stop()
+        super().tearDownClass()
+
+    @parameterized.expand(["train", "test"])
+    def test_mrpc(self, split):
+        dataset = MRPC(root=self.root_dir, split=split)
+
+        samples = list(dataset)
+        expected_samples = self.samples[split]
+        for sample, expected_sample in zip_equal(samples, expected_samples):
+            self.assertEqual(sample, expected_sample)
+
+    @parameterized.expand(["train", "test"])
+    def test_sst2_split_argument(self, split):
+        dataset1 = MRPC(root=self.root_dir, split=split)
+        (dataset2,) = MRPC(root=self.root_dir, split=(split,))
+
+        for d1, d2 in zip_equal(dataset1, dataset2):
+            self.assertEqual(d1, d2)

torchtext/datasets/__init__.py

@@ -11,6 +11,7 @@
 from .imdb import IMDB
 from .iwslt2016 import IWSLT2016
 from .iwslt2017 import IWSLT2017
+from .mrpc import MRPC
 from .multi30k import Multi30k
 from .penntreebank import PennTreebank
 from .sogounews import SogouNews
@@ -36,6 +37,7 @@
     "IMDB": IMDB,
     "IWSLT2016": IWSLT2016,
     "IWSLT2017": IWSLT2017,
+    "MRPC": MRPC,
     "Multi30k": Multi30k,
     "PennTreebank": PennTreebank,
     "SQuAD1": SQuAD1,

torchtext/datasets/mrpc.py

@@ -0,0 +1,73 @@
+import csv
+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, HttpReader, IterableWrapper
+
+
+URL = {
+    "train": "https://dl.fbaipublicfiles.com/senteval/senteval_data/msr_paraphrase_train.txt",
+    "test": "https://dl.fbaipublicfiles.com/senteval/senteval_data/msr_paraphrase_test.txt",
+}
+
+MD5 = {
+    "train": "793daf7b6224281e75fe61c1f80afe35",
+    "test": "e437fdddb92535b820fe8852e2df8a49",
+}
+
+NUM_LINES = {
+    "train": 4076,
+    "test": 1725,
+}
+
+
+DATASET_NAME = "MRPC"
+
+
+@_create_dataset_directory(dataset_name=DATASET_NAME)
+@_wrap_split_argument(("train", "test"))
+def MRPC(root: str, split: Union[Tuple[str], str]):
+    """MRPC Dataset
+
+    For additional details refer to https://www.microsoft.com/en-us/download/details.aspx?id=52398
+
+    Number of lines per split:
+        - train: 4076
+        - test: 1725
+
+    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 data points from MRPC dataset which consist of label, sentence1, sentence2
+    :rtype: (int, str, str)
+    """
+    if not is_module_available("torchdata"):
+        raise ModuleNotFoundError(
+            "Package `torchdata` not found. Please install following instructions at `https://github.com/pytorch/data`"
+        )
+
+    def _filepath_fn(x):
+        return os.path.join(root, os.path.basename(x))
+
+    def _modify_res(x):
+        return (int(x[0]), x[3], x[4])
+
+    url_dp = IterableWrapper([URL[split]])
+    # cache data on-disk with sanity check
+    cache_dp = url_dp.on_disk_cache(
+        filepath_fn=_filepath_fn,
+        hash_dict={_filepath_fn(URL[split]): MD5[split]},
+        hash_type="md5",
+    )
+    cache_dp = HttpReader(cache_dp).end_caching(mode="wb", same_filepath_fn=True)
+    cache_dp = FileOpener(cache_dp, encoding="utf-8")
+    cache_dp = cache_dp.parse_csv(skip_lines=1, delimiter="\t", quoting=csv.QUOTE_NONE).map(_modify_res)
+    return cache_dp.shuffle().set_shuffle(False).sharding_filter()