tensorflow · copybara-service · Jul 7, 2026 · Jul 7, 2026
diff --git a/tensorflow_datasets/datasets/stl10/stl10_dataset_builder.py b/tensorflow_datasets/datasets/stl10/stl10_dataset_builder.py
@@ -61,11 +61,11 @@ def gen_filenames(filenames):
     # Define the splits
     splits = [
         tfds.core.SplitGenerator(
-            name=tfds.Split.TRAIN,
+            name=tfds.Split.TRAIN,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={"filepaths": gen_filenames(train_files)},
         ),
         tfds.core.SplitGenerator(
-            name=tfds.Split.TEST,
+            name=tfds.Split.TEST,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={"filepaths": gen_filenames(test_files)},
         ),
         tfds.core.SplitGenerator(

diff --git a/tensorflow_datasets/datasets/story_cloze/story_cloze_dataset_builder.py b/tensorflow_datasets/datasets/story_cloze/story_cloze_dataset_builder.py
@@ -61,23 +61,23 @@ def _info(self) -> tfds.core.DatasetInfo:
 
   def _split_generators(self, dl_manager: tfds.download.DownloadManager):
     """Returns SplitGenerators."""
-    if self.builder_config.name == '2016':
+    if self.builder_config.name == '2016':  # pyrefly: ignore[missing-attribute]
       file_postfix = 'spring2016'
     elif self.builder_config.name == '2018':
       file_postfix = 'winter2018'
 
-    val_file = 'cloze_test_val__' + file_postfix + '.csv'
+    val_file = 'cloze_test_val__' + file_postfix + '.csv'  # pyrefly: ignore[unbound-name]
     test_file = 'cloze_test_test__' + file_postfix + '.csv'
 
     return [
         tfds.core.SplitGenerator(
-            name=tfds.Split.VALIDATION,
+            name=tfds.Split.VALIDATION,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={
                 'filepath': os.path.join(dl_manager.manual_dir, val_file)
             },
         ),
         tfds.core.SplitGenerator(
-            name=tfds.Split.TEST,
+            name=tfds.Split.TEST,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={
                 'filepath': os.path.join(dl_manager.manual_dir, test_file)
             },

diff --git a/tensorflow_datasets/datasets/sun397/sun397_dataset_builder.py b/tensorflow_datasets/datasets/sun397/sun397_dataset_builder.py
@@ -185,7 +185,7 @@ def __init__(self, tfds_split_files=None, **kwargs):
           "va": "sun397_tfds_va.txt",
       }
       for split, filename in tfds_split_files.items():
-        tfds_split_files[split] = tfds.core.tfds_path(
+        tfds_split_files[split] = tfds.core.tfds_path(  # pyrefly: ignore[bad-assignment]
             os.path.join("datasets", "sun397", filename)
         )
     self._tfds_split_files = tfds_split_files
@@ -222,25 +222,25 @@ def _split_generators(self, dl_manager):
         path=paths["images"],
         extract_method=tfds.download.ExtractMethod.TAR_GZ_STREAM,
     )
-    if self.builder_config.name == "tfds":
+    if self.builder_config.name == "tfds":  # pyrefly: ignore[missing-attribute]
       subset_images = self._get_tfds_subsets_images()
       return [
           tfds.core.SplitGenerator(
-              name=tfds.Split.TRAIN,
+              name=tfds.Split.TRAIN,  # pyrefly: ignore[missing-attribute]
               gen_kwargs=dict(
                   archive=dl_manager.iter_archive(images),
                   subset_images=subset_images["tr"],
               ),
           ),
           tfds.core.SplitGenerator(
-              name=tfds.Split.TEST,
+              name=tfds.Split.TEST,  # pyrefly: ignore[missing-attribute]
               gen_kwargs=dict(
                   archive=dl_manager.iter_archive(images),
                   subset_images=subset_images["te"],
               ),
           ),
           tfds.core.SplitGenerator(
-              name=tfds.Split.VALIDATION,
+              name=tfds.Split.VALIDATION,  # pyrefly: ignore[missing-attribute]
               gen_kwargs=dict(
                   archive=dl_manager.iter_archive(images),
                   subset_images=subset_images["va"],
@@ -251,14 +251,14 @@ def _split_generators(self, dl_manager):
       subset_images = self._get_partition_subsets_images(paths["partitions"])
       return [
           tfds.core.SplitGenerator(
-              name=tfds.Split.TRAIN,
+              name=tfds.Split.TRAIN,  # pyrefly: ignore[missing-attribute]
               gen_kwargs=dict(
                   archive=dl_manager.iter_archive(images),
                   subset_images=subset_images["tr"],
               ),
           ),
           tfds.core.SplitGenerator(
-              name=tfds.Split.TEST,
+              name=tfds.Split.TEST,  # pyrefly: ignore[missing-attribute]
               gen_kwargs=dict(
                   archive=dl_manager.iter_archive(images),
                   subset_images=subset_images["te"],
@@ -284,8 +284,8 @@ def _generate_examples(self, archive, subset_images):
                 fobj,
                 sess,
                 filename,
-                quality=self.builder_config.quality,
-                target_pixels=self.builder_config.target_pixels,
+                quality=self.builder_config.quality,  # pyrefly: ignore[missing-attribute]
+                target_pixels=self.builder_config.target_pixels,  # pyrefly: ignore[missing-attribute]
             )
             record = {
                 "file_name": filename,
@@ -307,8 +307,8 @@ def _get_partition_subsets_images(self, partitions_dir):
       all_images.update(split_images)
     # Load the images in the training/test split of this partition.
     filenames = {
-        "tr": "Training_%02d.txt" % self.builder_config.partition,
-        "te": "Testing_%02d.txt" % self.builder_config.partition,
+        "tr": "Training_%02d.txt" % self.builder_config.partition,  # pyrefly: ignore[missing-attribute]
+        "te": "Testing_%02d.txt" % self.builder_config.partition,  # pyrefly: ignore[missing-attribute]
     }
     splits_sets = {}
     for split, filename in filenames.items():

diff --git a/tensorflow_datasets/datasets/svhn_cropped/svhn_cropped_dataset_builder.py b/tensorflow_datasets/datasets/svhn_cropped/svhn_cropped_dataset_builder.py
@@ -40,7 +40,7 @@ def _info(self):
         "label": tfds.features.ClassLabel(num_classes=10),
     }
     if self.version > "3.0.0":
-      features_dict["id"] = tfds.features.Text()
+      features_dict["id"] = tfds.features.Text()  # pyrefly: ignore[bad-assignment]
     return self.dataset_info_from_configs(
         description=(
             "The Street View House Numbers (SVHN) Dataset is an image digit "
@@ -61,14 +61,14 @@ def _split_generators(self, dl_manager):
 
     return [
         tfds.core.SplitGenerator(
-            name=tfds.Split.TRAIN,
+            name=tfds.Split.TRAIN,  # pyrefly: ignore[missing-attribute]
             gen_kwargs=dict(
                 split_prefix="train_",
                 filepath=output_files["train"],
             ),
         ),
         tfds.core.SplitGenerator(
-            name=tfds.Split.TEST,
+            name=tfds.Split.TEST,  # pyrefly: ignore[missing-attribute]
             gen_kwargs=dict(
                 split_prefix="test_",
                 filepath=output_files["test"],

diff --git a/tensorflow_datasets/datasets/tatoeba/tatoeba_dataset_builder.py b/tensorflow_datasets/datasets/tatoeba/tatoeba_dataset_builder.py
@@ -103,7 +103,7 @@ def _info(self) -> tfds.core.DatasetInfo:
 
   def _split_generators(self, dl_manager: tfds.download.DownloadManager):
     """Returns SplitGenerators."""
-    lang = _LANGS[self.builder_config.language]
+    lang = _LANGS[self.builder_config.language]  # pyrefly: ignore[missing-attribute]
 
     tatoeba_source_path = os.path.join(
         _DATA_URLS, "tatoeba.{lang}-eng.{lang}".format(lang=lang)
@@ -138,6 +138,6 @@ def _generate_examples(self, source_file, target_file):
       yield i, {
           "source_sentence": source_sentences[i],
           "target_sentence": target_sentences[i],
-          "source_language": self.builder_config.language,
+          "source_language": self.builder_config.language,  # pyrefly: ignore[missing-attribute]
           "target_language": "en",
       }
diff --git a/tensorflow_datasets/datasets/ted_hrlr_translate/ted_hrlr_translate_dataset_builder.py b/tensorflow_datasets/datasets/ted_hrlr_translate/ted_hrlr_translate_dataset_builder.py
@@ -64,7 +64,7 @@ def __init__(self, *, language_pair=(None, None), **kwargs):
         first will be used as source and second as target in supervised mode.
       **kwargs: keyword arguments forwarded to super.
     """
-    name = "%s_to_%s" % (language_pair[0].replace("_", ""), language_pair[1])
+    name = "%s_to_%s" % (language_pair[0].replace("_", ""), language_pair[1])  # pyrefly: ignore[missing-attribute]
 
     description = "Translation dataset from %s to %s in plain text." % (
         language_pair[0],
@@ -99,21 +99,21 @@ class Builder(tfds.core.GeneratorBasedBuilder):
   def _info(self):
     return self.dataset_info_from_configs(
         features=tfds.features.Translation(
-            languages=self.builder_config.language_pair
+            languages=self.builder_config.language_pair  # pyrefly: ignore[missing-attribute]
         ),
         homepage="https://github.com/neulab/word-embeddings-for-nmt",
-        supervised_keys=self.builder_config.language_pair,
+        supervised_keys=self.builder_config.language_pair,  # pyrefly: ignore[missing-attribute]
     )
 
   def _split_generators(self, dl_manager):
     dl_dir = dl_manager.download_and_extract(_DATA_URL)
-    source, target = self.builder_config.language_pair
+    source, target = self.builder_config.language_pair  # pyrefly: ignore[missing-attribute]
 
     data_dir = os.path.join(dl_dir, "datasets", "%s_to_%s" % (source, target))
 
     return [
         tfds.core.SplitGenerator(
-            name=tfds.Split.TRAIN,
+            name=tfds.Split.TRAIN,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={
                 "source_file": os.path.join(
                     data_dir, "{}.train".format(source.replace("_", "-"))
@@ -124,7 +124,7 @@ def _split_generators(self, dl_manager):
             },
         ),
         tfds.core.SplitGenerator(
-            name=tfds.Split.VALIDATION,
+            name=tfds.Split.VALIDATION,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={
                 "source_file": os.path.join(
                     data_dir, "{}.dev".format(source.split("_")[0])
@@ -133,7 +133,7 @@ def _split_generators(self, dl_manager):
             },
         ),
         tfds.core.SplitGenerator(
-            name=tfds.Split.TEST,
+            name=tfds.Split.TEST,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={
                 "source_file": os.path.join(
                     data_dir, "{}.test".format(source.split("_")[0])
@@ -159,7 +159,7 @@ def _generate_examples(self, source_file, target_file):
         target_file,
     )
 
-    source, target = self.builder_config.language_pair
+    source, target = self.builder_config.language_pair  # pyrefly: ignore[missing-attribute]
     for idx, (l1, l2) in enumerate(zip(source_sentences, target_sentences)):
       result = {source: l1, target: l2}
       # Make sure that both translations are non-empty.

diff --git a/tensorflow_datasets/datasets/ted_multi_translate/ted_multi_translate_dataset_builder.py b/tensorflow_datasets/datasets/ted_multi_translate/ted_multi_translate_dataset_builder.py
@@ -108,17 +108,17 @@ def _split_generators(self, dl_manager):
 
     return [
         tfds.core.SplitGenerator(
-            name=tfds.Split.TRAIN,
+            name=tfds.Split.TRAIN,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={
                 'data_file': os.path.join(dl_dir, 'all_talks_train.tsv')
             },
         ),
         tfds.core.SplitGenerator(
-            name=tfds.Split.VALIDATION,
+            name=tfds.Split.VALIDATION,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={'data_file': os.path.join(dl_dir, 'all_talks_dev.tsv')},
         ),
         tfds.core.SplitGenerator(
-            name=tfds.Split.TEST,
+            name=tfds.Split.TEST,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={
                 'data_file': os.path.join(dl_dir, 'all_talks_test.tsv')
             },

diff --git a/tensorflow_datasets/datasets/tedlium/tedlium_dataset_builder.py b/tensorflow_datasets/datasets/tedlium/tedlium_dataset_builder.py
@@ -63,9 +63,9 @@ def _make_builder_configs():
       url="https://www.openslr.org/7/",
       download_url="http://www.openslr.org/resources/7/TEDLIUM_release1.tar.gz",
       split_paths=[
-          (tfds.Split.TRAIN, os.path.join("TEDLIUM_release1", "train")),
-          (tfds.Split.VALIDATION, os.path.join("TEDLIUM_release1", "dev")),
-          (tfds.Split.TEST, os.path.join("TEDLIUM_release1", "test")),
+          (tfds.Split.TRAIN, os.path.join("TEDLIUM_release1", "train")),  # pyrefly: ignore[missing-attribute]
+          (tfds.Split.VALIDATION, os.path.join("TEDLIUM_release1", "dev")),  # pyrefly: ignore[missing-attribute]
+          (tfds.Split.TEST, os.path.join("TEDLIUM_release1", "test")),  # pyrefly: ignore[missing-attribute]
       ],
   )
 
@@ -99,9 +99,9 @@ def _make_builder_configs():
           "http://www.openslr.org/resources/19/TEDLIUM_release2.tar.gz"
       ),
       split_paths=[
-          (tfds.Split.TRAIN, os.path.join("TEDLIUM_release2", "train")),
-          (tfds.Split.VALIDATION, os.path.join("TEDLIUM_release2", "dev")),
-          (tfds.Split.TEST, os.path.join("TEDLIUM_release2", "test")),
+          (tfds.Split.TRAIN, os.path.join("TEDLIUM_release2", "train")),  # pyrefly: ignore[missing-attribute]
+          (tfds.Split.VALIDATION, os.path.join("TEDLIUM_release2", "dev")),  # pyrefly: ignore[missing-attribute]
+          (tfds.Split.TEST, os.path.join("TEDLIUM_release2", "test")),  # pyrefly: ignore[missing-attribute]
       ],
   )
 
@@ -152,17 +152,17 @@ def _make_builder_configs():
       download_url="http://www.openslr.org/resources/51/TEDLIUM_release-3.tgz",
       split_paths=[
           (
-              tfds.Split.VALIDATION,
+              tfds.Split.VALIDATION,  # pyrefly: ignore[missing-attribute]
               os.path.join("TEDLIUM_release-3", "legacy", "dev"),
           ),
           (
-              tfds.Split.TEST,
+              tfds.Split.TEST,  # pyrefly: ignore[missing-attribute]
               os.path.join("TEDLIUM_release-3", "legacy", "test"),
           ),
           # The legacy/train directory contains symlinks to "data",
           # which are skipped by extraction (see above).
           # Work around this by manually dereferencing the links here.
-          (tfds.Split.TRAIN, os.path.join("TEDLIUM_release-3", "data")),
+          (tfds.Split.TRAIN, os.path.join("TEDLIUM_release-3", "data")),  # pyrefly: ignore[missing-attribute]
       ],
   )
 
@@ -186,19 +186,19 @@ def _info(self):
             "id": np.str_,
         }),
         supervised_keys=("speech", "text"),
-        homepage=self.builder_config.url,
-        citation=self.builder_config.citation,
+        homepage=self.builder_config.url,  # pyrefly: ignore[missing-attribute]
+        citation=self.builder_config.citation,  # pyrefly: ignore[missing-attribute]
         metadata=tfds.core.MetadataDict(
             sample_rate=16000,
         ),
     )
 
   def _split_generators(self, dl_manager):
     extracted_dir = dl_manager.download_and_extract(
-        self.builder_config.download_url
+        self.builder_config.download_url  # pyrefly: ignore[missing-attribute]
     )
     splits = []
-    for split, path in self.builder_config.split_paths:
+    for split, path in self.builder_config.split_paths:  # pyrefly: ignore[missing-attribute]
       kwargs = {"directory": os.path.join(extracted_dir, path)}
       splits.append(tfds.core.SplitGenerator(name=split, gen_kwargs=kwargs))
     return splits

diff --git a/tensorflow_datasets/datasets/tf_flowers/tf_flowers_dataset_builder.py b/tensorflow_datasets/datasets/tf_flowers/tf_flowers_dataset_builder.py
@@ -45,7 +45,7 @@ def _split_generators(self, dl_manager):
     # There is no predefined train/val/test split for this dataset.
     return [
         tfds.core.SplitGenerator(
-            name=tfds.Split.TRAIN,
+            name=tfds.Split.TRAIN,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={"images_dir_path": dl_manager.iter_archive(path)},
         ),
     ]

diff --git a/tensorflow_datasets/datasets/the300w_lp/the300w_lp_dataset_builder.py b/tensorflow_datasets/datasets/the300w_lp/the300w_lp_dataset_builder.py
@@ -66,7 +66,7 @@ def _split_generators(self, dl_manager):
     extracted_path = dl_manager.download_and_extract(_DATASET_URL)
     return [
         tfds.core.SplitGenerator(
-            name=tfds.Split.TRAIN,
+            name=tfds.Split.TRAIN,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={
                 "image_dir_path": os.path.join(extracted_path, "300W_LP"),
             },

diff --git a/tensorflow_datasets/datasets/tiny_shakespeare/tiny_shakespeare_dataset_builder.py b/tensorflow_datasets/datasets/tiny_shakespeare/tiny_shakespeare_dataset_builder.py
@@ -57,19 +57,19 @@ def _split_generators(self, dl_manager):
 
     return [
         tfds.core.SplitGenerator(
-            name=tfds.Split.TRAIN,
+            name=tfds.Split.TRAIN,  # pyrefly: ignore[missing-attribute]
             # These kwargs will be passed to _generate_examples
             gen_kwargs={'split_key': 'train', 'split_text': train_text},
         ),
         tfds.core.SplitGenerator(
-            name=tfds.Split.VALIDATION,
+            name=tfds.Split.VALIDATION,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={
                 'split_key': 'validation',
                 'split_text': validation_text,
             },
         ),
         tfds.core.SplitGenerator(
-            name=tfds.Split.TEST,
+            name=tfds.Split.TEST,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={'split_key': 'test', 'split_text': test_text},
         ),
     ]

diff --git a/tensorflow_datasets/datasets/titanic/titanic_dataset_builder.py b/tensorflow_datasets/datasets/titanic/titanic_dataset_builder.py
@@ -147,7 +147,7 @@ def _split_generators(self, dl_manager):
     # There is no predefined train/val/test split for this dataset.
     return [
         tfds.core.SplitGenerator(
-            name=tfds.Split.TRAIN, gen_kwargs={"file_path": path}
+            name=tfds.Split.TRAIN, gen_kwargs={"file_path": path}  # pyrefly: ignore[missing-attribute]
         ),
     ]
 

diff --git a/tensorflow_datasets/datasets/trec/trec_dataset_builder.py b/tensorflow_datasets/datasets/trec/trec_dataset_builder.py
@@ -102,13 +102,13 @@ def _split_generators(self, dl_manager):
     dl_files = dl_manager.download_and_extract(_URLs)
     return [
         tfds.core.SplitGenerator(
-            name=tfds.Split.TRAIN,
+            name=tfds.Split.TRAIN,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={
                 "filepath": dl_files["train"],
             },
         ),
         tfds.core.SplitGenerator(
-            name=tfds.Split.TEST,
+            name=tfds.Split.TEST,  # pyrefly: ignore[missing-attribute]
             gen_kwargs={
                 "filepath": dl_files["test"],
             },