Merge branch 'main' into wrappp

Author: NicolasHug

.github/scripts/setup-env.sh

@@ -40,9 +40,10 @@ conda create \
   --quiet --yes \
   python="${PYTHON_VERSION}" pip \
   ninja cmake \
-  libpng jpeg \
+  libpng \
   'ffmpeg<4.3'
 conda activate ci
+conda install --quiet --yes libjpeg-turbo -c pytorch
 pip install --progress-bar=off --upgrade setuptools
 
 # See https://github.com/pytorch/vision/issues/6790

.github/scripts/unittest.sh

@@ -11,4 +11,5 @@ echo '::group::Install testing utilities'
 pip install --progress-bar=off pytest pytest-mock pytest-cov
 echo '::endgroup::'
 
+python test/smoke_test.py
 pytest --junit-xml="${RUNNER_TEST_RESULTS_DIR}/test-results.xml" -v --durations=25

docs/source/transforms.rst

@@ -155,6 +155,7 @@ Color
 
     ColorJitter
     v2.ColorJitter
+    v2.RandomChannelPermutation
     v2.RandomPhotometricDistort
     Grayscale
     v2.Grayscale

packaging/pre_build_script.sh

@@ -13,8 +13,8 @@ fi
 
 if [[ "$(uname)" == Darwin || "$OSTYPE" == "msys" ]]; then
   # Install libpng from Anaconda (defaults)
-  conda install libpng "jpeg<=9b" -yq
-  conda install -yq ffmpeg=4.2 -c pytorch
+  conda install libpng -yq
+  conda install -yq ffmpeg=4.2 libjpeg-turbo -c pytorch
 
   # Copy binaries to be included in the wheel distribution
   if [[ "$OSTYPE" == "msys" ]]; then

packaging/torchvision/meta.yaml

@@ -10,7 +10,7 @@ requirements:
   build:
     - {{ compiler('c') }} # [win]
     - libpng
-    - jpeg
+    - libjpeg-turbo
     # NOTE: The only ffmpeg version that we build is actually 4.2
     - ffmpeg >=4.2  # [not win]
 
@@ -28,7 +28,7 @@ requirements:
     - requests
     - libpng
     - ffmpeg >=4.2  # [not win]
-    - jpeg
+    - libjpeg-turbo
     - pillow >=5.3.0, !=8.3.*
     - pytorch-mutex 1.0 {{ build_variant }}  # [not osx ]
     {{ environ.get('CONDA_PYTORCH_CONSTRAINT') }}
@@ -62,7 +62,7 @@ test:
   requires:
     - pytest
     - scipy
-    - jpeg
+    - libjpeg-turbo
     - ca-certificates
 
 

pytest.ini

@@ -3,7 +3,7 @@ addopts =
     # show tests that (f)ailed, (E)rror, or (X)passed in the summary
     -rfEX
     # Make tracebacks shorter
-    --tb=native
+    --tb=short
     # enable all warnings
     -Wd
     --ignore=test/test_datasets_download.py

test/common_utils.py

@@ -691,7 +691,7 @@ def make_bounding_box_loader(*, extra_dims=(), format, spatial_size=DEFAULT_PORT
     if isinstance(format, str):
         format = datapoints.BoundingBoxFormat[format]
 
-    spatial_size = _parse_size(spatial_size, name="canvas_size")
+    spatial_size = _parse_size(spatial_size, name="spatial_size")
 
     def fn(shape, dtype, device):
         *batch_dims, num_coordinates = shape
@@ -702,12 +702,12 @@ def fn(shape, dtype, device):
             format=format, canvas_size=spatial_size, batch_dims=batch_dims, dtype=dtype, device=device
         )
 
-    return BoundingBoxesLoader(fn, shape=(*extra_dims, 4), dtype=dtype, format=format, spatial_size=spatial_size)
+    return BoundingBoxesLoader(fn, shape=(*extra_dims[-1:], 4), dtype=dtype, format=format, spatial_size=spatial_size)
 
 
 def make_bounding_box_loaders(
     *,
-    extra_dims=DEFAULT_EXTRA_DIMS,
+    extra_dims=tuple(d for d in DEFAULT_EXTRA_DIMS if len(d) < 2),
     formats=tuple(datapoints.BoundingBoxFormat),
     spatial_size=DEFAULT_PORTRAIT_SPATIAL_SIZE,
     dtypes=(torch.float32, torch.float64, torch.int64),

test/smoke_test.py

@@ -78,6 +78,8 @@ def smoke_test_torchvision_resnet50_classify(device: str = "cpu") -> None:
 def main() -> None:
     print(f"torchvision: {torchvision.__version__}")
     print(f"torch.cuda.is_available: {torch.cuda.is_available()}")
+    print(f"{torch.ops.image._jpeg_version() = }")
+    assert torch.ops.image._is_compiled_against_turbo()
     smoke_test_torchvision()
     smoke_test_torchvision_read_decode()
     smoke_test_torchvision_resnet50_classify()

test/test_datapoints.py

@@ -22,7 +22,7 @@ def test_mask_instance(data):
     assert mask.ndim == 3 and mask.shape[0] == 1
 
 
-@pytest.mark.parametrize("data", [torch.randint(0, 32, size=(5, 4)), [[0, 0, 5, 5], [2, 2, 7, 7]]])
+@pytest.mark.parametrize("data", [torch.randint(0, 32, size=(5, 4)), [[0, 0, 5, 5], [2, 2, 7, 7]], [1, 2, 3, 4]])
 @pytest.mark.parametrize(
     "format", ["XYXY", "CXCYWH", datapoints.BoundingBoxFormat.XYXY, datapoints.BoundingBoxFormat.XYWH]
 )
@@ -35,6 +35,12 @@ def test_bbox_instance(data, format):
     assert bboxes.format == format
 
 
+def test_bbox_dim_error():
+    data_3d = [[[1, 2, 3, 4]]]
+    with pytest.raises(ValueError, match="Expected a 1D or 2D tensor, got 3D"):
+        datapoints.BoundingBoxes(data_3d, format="XYXY", canvas_size=(32, 32))
+
+
 @pytest.mark.parametrize(
     ("data", "input_requires_grad", "expected_requires_grad"),
     [

test/test_prototype_transforms.py

@@ -20,7 +20,7 @@
 from torchvision.datapoints import BoundingBoxes, BoundingBoxFormat, Image, Mask, Video
 from torchvision.prototype import datapoints, transforms
 from torchvision.transforms.v2._utils import _convert_fill_arg
-from torchvision.transforms.v2.functional import InterpolationMode, pil_to_tensor, to_image_pil
+from torchvision.transforms.v2.functional import clamp_bounding_boxes, InterpolationMode, pil_to_tensor, to_image_pil
 from torchvision.transforms.v2.utils import check_type, is_simple_tensor
 
 BATCH_EXTRA_DIMS = [extra_dims for extra_dims in DEFAULT_EXTRA_DIMS if extra_dims]
@@ -306,7 +306,9 @@ def test__transform_bounding_boxes_clamping(self, mocker):
         bounding_boxes = make_bounding_box(
             format=BoundingBoxFormat.XYXY, canvas_size=canvas_size, batch_dims=(batch_size,)
         )
-        mock = mocker.patch("torchvision.prototype.transforms._geometry.F.clamp_bounding_boxes")
+        mock = mocker.patch(
+            "torchvision.prototype.transforms._geometry.F.clamp_bounding_boxes", wraps=clamp_bounding_boxes
+        )
 
         transform = transforms.FixedSizeCrop((-1, -1))
         mocker.patch("torchvision.prototype.transforms._geometry.has_any", return_value=True)