Migrating AmazonReviewPolarity to datapipes

test/data/test_builtin_datasets.py

@@ -38,20 +38,6 @@ def test_raw_ag_news(self):
         self._helper_test_func(len(test_iter), 7600, next(test_iter)[1][:25], 'Fears for T N pension aft')
         del train_iter, test_iter
 
-    @parameterized.expand(
-        load_params('raw_datasets.jsonl'),
-        name_func=_raw_text_custom_name_func)
-    def test_raw_text_name_property(self, info):
-        dataset_name = info['dataset_name']
-        split = info['split']
-
-        if dataset_name == 'WMT14':
-            return
-        else:
-            data_iter = torchtext.datasets.DATASETS[dataset_name](split=split)
-
-        self.assertEqual(str(data_iter), dataset_name)
-
     @parameterized.expand(
         load_params('raw_datasets.jsonl'),
         name_func=_raw_text_custom_name_func)
@@ -63,8 +49,7 @@ def test_raw_text_classification(self, info):
             return
         else:
             data_iter = torchtext.datasets.DATASETS[dataset_name](split=split)
-        self.assertEqual(len(data_iter), info['NUM_LINES'])
-        self.assertEqual(hashlib.md5(json.dumps(next(data_iter), sort_keys=True).encode('utf-8')).hexdigest(), info['first_line'])
+        self.assertEqual(hashlib.md5(json.dumps(next(iter(data_iter)), sort_keys=True).encode('utf-8')).hexdigest(), info['first_line'])
         if dataset_name == "AG_NEWS":
             self.assertEqual(torchtext.datasets.URLS[dataset_name][split], info['URL'])
             self.assertEqual(torchtext.datasets.MD5[dataset_name][split], info['MD5'])

torchtext/data/datasets_utils.py

@@ -209,8 +209,7 @@ def _wrap_split_argument_with_fn(fn, splits):
             argspec.args[1] == "split" and
             argspec.varargs is None and
             argspec.varkw is None and
-            len(argspec.kwonlyargs) == 0 and
-            len(argspec.annotations) == 0
+            len(argspec.kwonlyargs) == 0
             ):
         raise ValueError("Internal Error: Given function {} did not adhere to standard signature.".format(fn))
 
@@ -240,23 +239,22 @@ def new_fn(fn):
 
 
 def _create_dataset_directory(dataset_name):
-    def decorator(func):
-        argspec = inspect.getfullargspec(func)
+    def decorator(fn):
+        argspec = inspect.getfullargspec(fn)
         if not (argspec.args[0] == "root" and
                 argspec.args[1] == "split" and
                 argspec.varargs is None and
                 argspec.varkw is None and
-                len(argspec.kwonlyargs) == 0 and
-                len(argspec.annotations) == 0
+                len(argspec.kwonlyargs) == 0
                 ):
             raise ValueError("Internal Error: Given function {} did not adhere to standard signature.".format(fn))
 
-        @functools.wraps(func)
+        @functools.wraps(fn)
         def wrapper(root=_CACHE_DIR, *args, **kwargs):
             new_root = os.path.join(root, dataset_name)
             if not os.path.exists(new_root):
                 os.makedirs(new_root)
-            return func(root=new_root, *args, **kwargs)
+            return fn(root=new_root, *args, **kwargs)
 
         return wrapper
 

torchtext/datasets/amazonreviewpolarity.py

@@ -1,13 +1,15 @@
+from torchtext._internal.module_utils import is_module_available
+from typing import Union, Tuple
+if is_module_available("torchdata"):
+    from torchdata.datapipes.iter import FileOpener, GDriveReader, IterableWrapper
+
 from torchtext.data.datasets_utils import (
-    _RawTextIterableDataset,
     _wrap_split_argument,
     _add_docstring_header,
-    _download_extract_validate,
     _create_dataset_directory,
-    _create_data_from_csv,
 )
+
 import os
-import logging
 
 URL = 'https://drive.google.com/uc?export=download&id=0Bz8a_Dbh9QhbaW12WVVZS2drcnM'
 
@@ -25,20 +27,24 @@
     'test': f'{os.sep}'.join(['amazon_review_polarity_csv', 'test.csv']),
 }
 
-_EXTRACTED_FILES_MD5 = {
-    'train': "520937107c39a2d1d1f66cd410e9ed9e",
-    'test': "f4c8bded2ecbde5f996b675db6228f16"
-}
 
 DATASET_NAME = "AmazonReviewPolarity"
 
 
 @_add_docstring_header(num_lines=NUM_LINES, num_classes=2)
 @_create_dataset_directory(dataset_name=DATASET_NAME)
-@_wrap_split_argument(('train', 'test'))
-def AmazonReviewPolarity(root, split):
-    path = _download_extract_validate(root, URL, MD5, os.path.join(root, _PATH), os.path.join(root, _EXTRACTED_FILES[split]),
-                                      _EXTRACTED_FILES_MD5[split], hash_type="md5")
-    logging.info('Creating {} data'.format(split))
-    return _RawTextIterableDataset(DATASET_NAME, NUM_LINES[split],
-                                   _create_data_from_csv(path))
+@_wrap_split_argument(("train", "test"))
+def AmazonReviewPolarity(root: str, split: Union[Tuple[str], str]):
+    # TODO Remove this after removing conditional dependency
+    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_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_dp = GDriveReader(cache_dp).end_caching(mode="wb", same_filepath_fn=True)
+    cache_dp = FileOpener(cache_dp, mode="b")
+    extracted_files = cache_dp.read_from_tar()
+    filter_extracted_files = extracted_files.filter(lambda x: split in x[0])
+    return filter_extracted_files.parse_csv().map(fn=lambda t: (int(t[0]), ' '.join(t[1:])))