fix: forbid bool value from JPEG's quality argument type. (pytorch#8995)

Co-authored-by: Nicolas Hug <[email protected]>

Author: GdoongMathew

test/test_transforms_v2.py

@@ -6192,6 +6192,11 @@ def test_transform_invalid_quality_error(self, quality):
         with pytest.raises(ValueError, match="quality must be an integer from 1 to 100"):
             transforms.JPEG(quality=quality)
 
+    @pytest.mark.parametrize("quality", [None, True])
+    def test_transform_quality_type_error(self, quality):
+        with pytest.raises(TypeError, match="quality"):
+            transforms.JPEG(quality=quality)
+
 
 class TestUtils:
     # TODO: Still need to test has_all, has_any, check_type and get_bouding_boxes

torchvision/transforms/v2/_augment.py

@@ -352,6 +352,8 @@ class JPEG(Transform):
     def __init__(self, quality: Union[int, Sequence[int]]):
         super().__init__()
         if isinstance(quality, int):
+            if isinstance(quality, bool):
+                raise TypeError("quality can't be bool")
             quality = [quality, quality]
         else:
             _check_sequence_input(quality, "quality", req_sizes=(2,))