Merge branch 'main' into issue-8203

.github/workflows/build-wheels-aarch64-linux.yml

@@ -13,6 +13,10 @@ on:
         - v[0-9]+.[0-9]+.[0-9]+-rc[0-9]+
   workflow_dispatch:
 
+permissions:
+  id-token: write
+  contents: read
+
 jobs:
   generate-matrix:
     uses: pytorch/test-infra/.github/workflows/generate_binary_build_matrix.yml@main
@@ -48,6 +52,3 @@ jobs:
       trigger-event: ${{ github.event_name }}
       architecture: aarch64
       setup-miniconda: false
-    secrets:
-      AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID: ${{ secrets.AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID }}
-      AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY: ${{ secrets.AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY }}

.github/workflows/build-wheels-linux.yml

@@ -13,6 +13,10 @@ on:
         - v[0-9]+.[0-9]+.[0-9]+-rc[0-9]+
   workflow_dispatch:
 
+permissions:
+  id-token: write
+  contents: read
+
 jobs:
   generate-matrix:
     uses: pytorch/test-infra/.github/workflows/generate_binary_build_matrix.yml@main
@@ -45,6 +49,3 @@ jobs:
       package-name: ${{ matrix.package-name }}
       smoke-test-script: ${{ matrix.smoke-test-script }}
       trigger-event: ${{ github.event_name }}
-    secrets:
-      AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID: ${{ secrets.AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID }}
-      AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY: ${{ secrets.AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY }}

.github/workflows/build-wheels-m1.yml

@@ -13,6 +13,10 @@ on:
         - v[0-9]+.[0-9]+.[0-9]+-rc[0-9]+
   workflow_dispatch:
 
+permissions:
+  id-token: write
+  contents: read
+
 jobs:
   generate-matrix:
     uses: pytorch/test-infra/.github/workflows/generate_binary_build_matrix.yml@main
@@ -46,6 +50,3 @@ jobs:
       runner-type: macos-m1-12
       smoke-test-script: ${{ matrix.smoke-test-script }}
       trigger-event: ${{ github.event_name }}
-    secrets:
-      AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID: ${{ secrets.AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID }}
-      AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY: ${{ secrets.AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY }}

.github/workflows/build-wheels-macos.yml

@@ -13,6 +13,10 @@ on:
         - v[0-9]+.[0-9]+.[0-9]+-rc[0-9]+
   workflow_dispatch:
 
+permissions:
+  id-token: write
+  contents: read
+
 jobs:
   generate-matrix:
     uses: pytorch/test-infra/.github/workflows/generate_binary_build_matrix.yml@main
@@ -46,6 +50,3 @@ jobs:
       runner-type: macos-12
       smoke-test-script: ${{ matrix.smoke-test-script }}
       trigger-event: ${{ github.event_name }}
-    secrets:
-      AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID: ${{ secrets.AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID }}
-      AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY: ${{ secrets.AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY }}

.github/workflows/build-wheels-windows.yml

@@ -13,6 +13,10 @@ on:
         - v[0-9]+.[0-9]+.[0-9]+-rc[0-9]+
   workflow_dispatch:
 
+permissions:
+  id-token: write
+  contents: read
+
 jobs:
   generate-matrix:
     uses: pytorch/test-infra/.github/workflows/generate_binary_build_matrix.yml@main
@@ -47,6 +51,3 @@ jobs:
       package-name: ${{ matrix.package-name }}
       smoke-test-script: ${{ matrix.smoke-test-script }}
       trigger-event: ${{ github.event_name }}
-    secrets:
-      AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID: ${{ secrets.AWS_PYTORCH_UPLOADER_ACCESS_KEY_ID }}
-      AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY: ${{ secrets.AWS_PYTORCH_UPLOADER_SECRET_ACCESS_KEY }}

.github/workflows/pr-labels.yml

@@ -21,7 +21,7 @@ jobs:
 
       - name: Process commit and find merger responsible for labeling
         id: commit
-        run: echo "::set-output name=merger::$(python .github/process_commit.py ${{ github.sha }})"
+        run: echo "merger=$(python .github/process_commit.py ${{ github.sha }})" >> $GITHUB_OUTPUT
 
       - name: Ping merger responsible for labeling if necessary
         if: ${{ steps.commit.outputs.merger != '' }}

CONTRIBUTING.md

@@ -34,7 +34,9 @@ clear and has sufficient instructions to be able to reproduce the issue.
 ### Dependencies
 
 Start by installing the **nightly** build of PyTorch following the [official
-instructions](https://pytorch.org/get-started/locally/).
+instructions](https://pytorch.org/get-started/locally/). Note that the official
+instructions may ask you to install torchvision itself. If you are doing development
+on torchvision, you should not install prebuilt torchvision packages.
 
 **Optionally**, install `libpng` and `libjpeg-turbo` if you want to enable
 support for
@@ -69,10 +71,10 @@ building GPU support by setting `FORCE_CUDA=1` environment variable, which is us
 We don't officially support building from source using `pip`, but _if_ you do, you'll need to use the
 `--no-build-isolation` flag.
 
-Other development dependencies include:
+#### Other development dependencies (some of these are needed to run tests):
 
 ```
-pip install flake8 typing mypy pytest pytest-mock scipy
+pip install expecttest flake8 typing mypy pytest pytest-mock scipy
 ```
 
 ## Development Process
@@ -145,6 +147,8 @@ mypy --config-file mypy.ini
 
 ### Unit tests
 
+Before running tests make sure to install [test dependencies](#other-development-dependencies-some-of-these-are-needed-to-run-tests).
+
 If you have modified the code by adding a new feature or a bug-fix, please add unit tests for that. To run a specific
 test:
 ```bash

docs/source/models/ssdlite.rst

@@ -6,7 +6,7 @@ SSDlite
 The SSDLite model is based on the `SSD: Single Shot MultiBox Detector
 <https://arxiv.org/abs/1512.02325>`__, `Searching for MobileNetV3
 <https://arxiv.org/abs/1905.02244>`__ and `MobileNetV2: Inverted Residuals and Linear
-Bottlenecks <https://arxiv.org/abs/1801.04381>__` papers.
+Bottlenecks <https://arxiv.org/abs/1801.04381>`__ papers.
 
 .. betastatus:: detection module
 

test/test_datasets_utils.py

@@ -58,8 +58,11 @@ def test_get_redirect_url_max_hops_exceeded(self, mocker):
         assert mock.call_count == 1
         assert mock.call_args[0][0].full_url == url
 
-    def test_check_md5(self):
+    @pytest.mark.parametrize("use_pathlib", (True, False))
+    def test_check_md5(self, use_pathlib):
         fpath = TEST_FILE
+        if use_pathlib:
+            fpath = pathlib.Path(fpath)
         correct_md5 = "9c0bb82894bb3af7f7675ef2b3b6dcdc"
         false_md5 = ""
         assert utils.check_md5(fpath, correct_md5)
@@ -116,7 +119,8 @@ def test_detect_file_type_incompatible(self, file):
             utils._detect_file_type(file)
 
     @pytest.mark.parametrize("extension", [".bz2", ".gz", ".xz"])
-    def test_decompress(self, extension, tmpdir):
+    @pytest.mark.parametrize("use_pathlib", (True, False))
+    def test_decompress(self, extension, tmpdir, use_pathlib):
         def create_compressed(root, content="this is the content"):
             file = os.path.join(root, "file")
             compressed = f"{file}{extension}"
@@ -128,6 +132,8 @@ def create_compressed(root, content="this is the content"):
             return compressed, file, content
 
         compressed, file, content = create_compressed(tmpdir)
+        if use_pathlib:
+            compressed = pathlib.Path(compressed)
 
         utils._decompress(compressed)
 
@@ -140,7 +146,8 @@ def test_decompress_no_compression(self):
         with pytest.raises(RuntimeError):
             utils._decompress("foo.tar")
 
-    def test_decompress_remove_finished(self, tmpdir):
+    @pytest.mark.parametrize("use_pathlib", (True, False))
+    def test_decompress_remove_finished(self, tmpdir, use_pathlib):
         def create_compressed(root, content="this is the content"):
             file = os.path.join(root, "file")
             compressed = f"{file}.gz"
@@ -151,10 +158,20 @@ def create_compressed(root, content="this is the content"):
             return compressed, file, content
 
         compressed, file, content = create_compressed(tmpdir)
+        print(f"{type(compressed)=}")
+        if use_pathlib:
+            compressed = pathlib.Path(compressed)
+            tmpdir = pathlib.Path(tmpdir)
 
-        utils.extract_archive(compressed, tmpdir, remove_finished=True)
+        extracted_dir = utils.extract_archive(compressed, tmpdir, remove_finished=True)
 
         assert not os.path.exists(compressed)
+        if use_pathlib:
+            assert isinstance(extracted_dir, pathlib.Path)
+            assert isinstance(compressed, pathlib.Path)
+        else:
+            assert isinstance(extracted_dir, str)
+            assert isinstance(compressed, str)
 
     @pytest.mark.parametrize("extension", [".gz", ".xz"])
     @pytest.mark.parametrize("remove_finished", [True, False])
@@ -167,7 +184,8 @@ def test_extract_archive_defer_to_decompress(self, extension, remove_finished, m
 
         mocked.assert_called_once_with(file, filename, remove_finished=remove_finished)
 
-    def test_extract_zip(self, tmpdir):
+    @pytest.mark.parametrize("use_pathlib", (True, False))
+    def test_extract_zip(self, tmpdir, use_pathlib):
         def create_archive(root, content="this is the content"):
             file = os.path.join(root, "dst.txt")
             archive = os.path.join(root, "archive.zip")
@@ -177,6 +195,8 @@ def create_archive(root, content="this is the content"):
 
             return archive, file, content
 
+        if use_pathlib:
+            tmpdir = pathlib.Path(tmpdir)
         archive, file, content = create_archive(tmpdir)
 
         utils.extract_archive(archive, tmpdir)
@@ -189,7 +209,8 @@ def create_archive(root, content="this is the content"):
     @pytest.mark.parametrize(
         "extension, mode", [(".tar", "w"), (".tar.gz", "w:gz"), (".tgz", "w:gz"), (".tar.xz", "w:xz")]
     )
-    def test_extract_tar(self, extension, mode, tmpdir):
+    @pytest.mark.parametrize("use_pathlib", (True, False))
+    def test_extract_tar(self, extension, mode, tmpdir, use_pathlib):
         def create_archive(root, extension, mode, content="this is the content"):
             src = os.path.join(root, "src.txt")
             dst = os.path.join(root, "dst.txt")
@@ -203,6 +224,8 @@ def create_archive(root, extension, mode, content="this is the content"):
 
             return archive, dst, content
 
+        if use_pathlib:
+            tmpdir = pathlib.Path(tmpdir)
         archive, file, content = create_archive(tmpdir, extension, mode)
 
         utils.extract_archive(archive, tmpdir)

test/test_internet.py

@@ -6,35 +6,48 @@
 """
 
 import os
+import pathlib
 from urllib.error import URLError
 
 import pytest
 import torchvision.datasets.utils as utils
 
 
 class TestDatasetUtils:
-    def test_download_url(self, tmpdir):
+    @pytest.mark.parametrize("use_pathlib", (True, False))
+    def test_download_url(self, tmpdir, use_pathlib):
+        if use_pathlib:
+            tmpdir = pathlib.Path(tmpdir)
         url = "http://github.com/pytorch/vision/archive/master.zip"
         try:
             utils.download_url(url, tmpdir)
             assert len(os.listdir(tmpdir)) != 0
         except URLError:
             pytest.skip(f"could not download test file '{url}'")
 
-    def test_download_url_retry_http(self, tmpdir):
+    @pytest.mark.parametrize("use_pathlib", (True, False))
+    def test_download_url_retry_http(self, tmpdir, use_pathlib):
+        if use_pathlib:
+            tmpdir = pathlib.Path(tmpdir)
         url = "https://github.com/pytorch/vision/archive/master.zip"
         try:
             utils.download_url(url, tmpdir)
             assert len(os.listdir(tmpdir)) != 0
         except URLError:
             pytest.skip(f"could not download test file '{url}'")
 
-    def test_download_url_dont_exist(self, tmpdir):
+    @pytest.mark.parametrize("use_pathlib", (True, False))
+    def test_download_url_dont_exist(self, tmpdir, use_pathlib):
+        if use_pathlib:
+            tmpdir = pathlib.Path(tmpdir)
         url = "http://github.com/pytorch/vision/archive/this_doesnt_exist.zip"
         with pytest.raises(URLError):
             utils.download_url(url, tmpdir)
 
-    def test_download_url_dispatch_download_from_google_drive(self, mocker, tmpdir):
+    @pytest.mark.parametrize("use_pathlib", (True, False))
+    def test_download_url_dispatch_download_from_google_drive(self, mocker, tmpdir, use_pathlib):
+        if use_pathlib:
+            tmpdir = pathlib.Path(tmpdir)
         url = "https://drive.google.com/file/d/1GO-BHUYRuvzr1Gtp2_fqXRsr9TIeYbhV/view"
 
         id = "1GO-BHUYRuvzr1Gtp2_fqXRsr9TIeYbhV"
@@ -44,7 +57,7 @@ def test_download_url_dispatch_download_from_google_drive(self, mocker, tmpdir):
         mocked = mocker.patch("torchvision.datasets.utils.download_file_from_google_drive")
         utils.download_url(url, tmpdir, filename, md5)
 
-        mocked.assert_called_once_with(id, tmpdir, filename, md5)
+        mocked.assert_called_once_with(id, os.path.expanduser(tmpdir), filename, md5)
 
 
 if __name__ == "__main__":

test/test_utils.py

@@ -210,14 +210,11 @@ def test_draw_segmentation_masks(colors, alpha, device):
     num_masks, h, w = 2, 100, 100
     dtype = torch.uint8
     img = torch.randint(0, 256, size=(3, h, w), dtype=dtype, device=device)
-    masks = torch.randint(0, 2, (num_masks, h, w), dtype=torch.bool, device=device)
+    masks = torch.zeros((num_masks, h, w), dtype=torch.bool, device=device)
+    masks[0, 10:20, 10:20] = True
+    masks[1, 15:25, 15:25] = True
 
-    # For testing we enforce that there's no overlap between the masks. The
-    # current behaviour is that the last mask's color will take priority when
-    # masks overlap, but this makes testing slightly harder, so we don't really
-    # care
     overlap = masks[0] & masks[1]
-    masks[:, overlap] = False
 
     out = utils.draw_segmentation_masks(img, masks, colors=colors, alpha=alpha)
     assert out.dtype == dtype
@@ -239,12 +236,15 @@ def test_draw_segmentation_masks(colors, alpha, device):
         color = torch.tensor(color, dtype=dtype, device=device)
 
         if alpha == 1:
-            assert (out[:, mask] == color[:, None]).all()
+            assert (out[:, mask & ~overlap] == color[:, None]).all()
         elif alpha == 0:
-            assert (out[:, mask] == img[:, mask]).all()
+            assert (out[:, mask & ~overlap] == img[:, mask & ~overlap]).all()
 
-        interpolated_color = (img[:, mask] * (1 - alpha) + color[:, None] * alpha).to(dtype)
-        torch.testing.assert_close(out[:, mask], interpolated_color, rtol=0.0, atol=1.0)
+        interpolated_color = (img[:, mask & ~overlap] * (1 - alpha) + color[:, None] * alpha).to(dtype)
+        torch.testing.assert_close(out[:, mask & ~overlap], interpolated_color, rtol=0.0, atol=1.0)
+
+    interpolated_overlap = (img[:, overlap] * (1 - alpha)).to(dtype)
+    torch.testing.assert_close(out[:, overlap], interpolated_overlap, rtol=0.0, atol=1.0)
 
 
 def test_draw_segmentation_masks_dtypes():

torchvision/csrc/io/decoder/stream.cpp

@@ -63,15 +63,8 @@ int Stream::openCodec(std::vector<DecoderMetadata>* metadata, int num_threads) {
     codecCtx_->thread_count = num_threads;
   } else {
     // otherwise set sensible defaults
-    // with the special case for the different MPEG4 codecs
-    // that don't have threading context functions
-    if (codecCtx_->codec->capabilities & AV_CODEC_CAP_INTRA_ONLY) {
-      codecCtx_->thread_type = FF_THREAD_FRAME;
-      codecCtx_->thread_count = 2;
-    } else {
-      codecCtx_->thread_count = 8;
-      codecCtx_->thread_type = FF_THREAD_SLICE;
-    }
+    codecCtx_->thread_count = 8;
+    codecCtx_->thread_type = FF_THREAD_SLICE;
   }
 
   int ret;

torchvision/datasets/caltech.py

@@ -23,7 +23,7 @@ class Caltech101(VisionDataset):
             target types.  ``category`` represents the target class, and
             ``annotation`` is a list of points from a hand-generated outline.
             Defaults to ``category``.
-        transform (callable, optional): A function/transform that takes in an PIL image
+        transform (callable, optional): A function/transform that takes in a PIL image
             and returns a transformed version. E.g, ``transforms.RandomCrop``
         target_transform (callable, optional): A function/transform that takes in the
             target and transforms it.
@@ -151,7 +151,7 @@ class Caltech256(VisionDataset):
     Args:
         root (string): Root directory of dataset where directory
             ``caltech256`` exists or will be saved to if download is set to True.
-        transform (callable, optional): A function/transform that takes in an PIL image
+        transform (callable, optional): A function/transform that takes in a PIL image
             and returns a transformed version. E.g, ``transforms.RandomCrop``
         target_transform (callable, optional): A function/transform that takes in the
             target and transforms it.

torchvision/datasets/celeba.py

@@ -31,7 +31,7 @@ class CelebA(VisionDataset):
 
             Defaults to ``attr``. If empty, ``None`` will be returned as target.
 
-        transform (callable, optional): A function/transform that  takes in an PIL image
+        transform (callable, optional): A function/transform that takes in a PIL image
             and returns a transformed version. E.g, ``transforms.PILToTensor``
         target_transform (callable, optional): A function/transform that takes in the
             target and transforms it.

torchvision/datasets/cifar.py

@@ -17,7 +17,7 @@ class CIFAR10(VisionDataset):
             ``cifar-10-batches-py`` exists or will be saved to if download is set to True.
         train (bool, optional): If True, creates dataset from training set, otherwise
             creates from test set.
-        transform (callable, optional): A function/transform that takes in an PIL image
+        transform (callable, optional): A function/transform that takes in a PIL image
             and returns a transformed version. E.g, ``transforms.RandomCrop``
         target_transform (callable, optional): A function/transform that takes in the
             target and transforms it.