Merge branch 'main' into patch-1

Author: NicolasHug

.github/scripts/setup-env.sh

@@ -45,11 +45,7 @@ echo '::endgroup::'
 
 if [[ "${OS_TYPE}" == windows && "${GPU_ARCH_TYPE}" == cuda ]]; then
   echo '::group::Install VisualStudio CUDA extensions on Windows'
-  if [[ "${VC_YEAR:-}" == "2022" ]]; then
-    TARGET_DIR="/c/Program Files (x86)/Microsoft Visual Studio/2022/BuildTools/MSBuild/Microsoft/VC/v170/BuildCustomizations"
-  else
-    TARGET_DIR="/c/Program Files (x86)/Microsoft Visual Studio/2019/BuildTools/MSBuild/Microsoft/VC/v160/BuildCustomizations"
-  fi
+  TARGET_DIR="/c/Program Files (x86)/Microsoft Visual Studio/2022/BuildTools/MSBuild/Microsoft/VC/v170/BuildCustomizations"
   mkdir -p "${TARGET_DIR}"
   cp -r "${CUDA_HOME}/MSBuildExtensions/"* "${TARGET_DIR}"
   echo '::endgroup::'

.github/workflows/tests.yml

@@ -102,7 +102,7 @@ jobs:
         set -euxo pipefail
 
         export PYTHON_VERSION=${{ matrix.python-version }}
-        export VC_YEAR=2019
+        export VC_YEAR=2022
         export VSDEVCMD_ARGS=""
         export GPU_ARCH_TYPE=${{ matrix.gpu-arch-type }}
         export GPU_ARCH_VERSION=${{ matrix.gpu-arch-version }}

packaging/windows/internal/vc_env_helper.bat

@@ -2,14 +2,6 @@
 
 set VC_VERSION_LOWER=17
 set VC_VERSION_UPPER=18
-if "%VC_YEAR%" == "2019" (
-    set VC_VERSION_LOWER=16
-    set VC_VERSION_UPPER=17
-)
-if "%VC_YEAR%" == "2017" (
-    set VC_VERSION_LOWER=15
-    set VC_VERSION_UPPER=16
-)
 
 for /f "usebackq tokens=*" %%i in (`"%ProgramFiles(x86)%\Microsoft Visual Studio\Installer\vswhere.exe" -legacy -products * -version [%VC_VERSION_LOWER%^,%VC_VERSION_UPPER%^) -property installationPath`) do (
     if exist "%%i" if exist "%%i\VC\Auxiliary\Build\vcvarsall.bat" (

setup.py

@@ -79,7 +79,7 @@ def get_version():
 
 def write_version_file(version, sha):
     # Exists for BC, probably completely useless.
-    with open(ROOT_DIR / "torchvision/version.py", "w") as f:
+    with open(ROOT_DIR / "torchvision" / "version.py", "w") as f:
         f.write(f"__version__ = '{version}'\n")
         f.write(f"git_version = {repr(sha)}\n")
         f.write("from torchvision.extension import _check_cuda_version\n")
@@ -194,7 +194,7 @@ def make_C_extension():
 
 def find_libpng():
     # Returns (found, include dir, library dir, library name)
-    if sys.platform in ("linux", "darwin"):
+    if sys.platform in ("linux", "darwin", "aix"):
         libpng_config = shutil.which("libpng-config")
         if libpng_config is None:
             warnings.warn("libpng-config not found")

test/test_datasets.py

@@ -532,7 +532,8 @@ def inject_fake_data(self, tmpdir, config):
         self._create_bbox_txt(base_folder, num_images)
         self._create_landmarks_txt(base_folder, num_images)
 
-        return dict(num_examples=num_images_per_split[config["split"]], attr_names=attr_names)
+        num_samples = num_images_per_split.get(config["split"], 0) if isinstance(config["split"], str) else 0
+        return dict(num_examples=num_samples, attr_names=attr_names)
 
     def _create_split_txt(self, root):
         num_images_per_split = dict(train=4, valid=3, test=2)
@@ -635,6 +636,28 @@ def test_transforms_v2_wrapper_spawn(self):
             with self.create_dataset(target_type=target_type, transform=v2.Resize(size=expected_size)) as (dataset, _):
                 datasets_utils.check_transforms_v2_wrapper_spawn(dataset, expected_size=expected_size)
 
+    def test_invalid_split_list(self):
+        with pytest.raises(ValueError, match="Expected type str for argument split, but got type <class 'list'>."):
+            with self.create_dataset(split=[1]):
+                pass
+
+    def test_invalid_split_int(self):
+        with pytest.raises(ValueError, match="Expected type str for argument split, but got type <class 'int'>."):
+            with self.create_dataset(split=1):
+                pass
+
+    def test_invalid_split_value(self):
+        with pytest.raises(
+            ValueError,
+            match="Unknown value '{value}' for argument {arg}. Valid values are {{{valid_values}}}.".format(
+                value="invalid",
+                arg="split",
+                valid_values=("train", "valid", "test", "all"),
+            ),
+        ):
+            with self.create_dataset(split="invalid"):
+                pass
+
 
 class VOCSegmentationTestCase(datasets_utils.ImageDatasetTestCase):
     DATASET_CLASS = datasets.VOCSegmentation

torchvision/datasets/celeba.py

@@ -93,7 +93,13 @@ def __init__(
             "test": 2,
             "all": None,
         }
-        split_ = split_map[verify_str_arg(split.lower(), "split", ("train", "valid", "test", "all"))]
+        split_ = split_map[
+            verify_str_arg(
+                split.lower() if isinstance(split, str) else split,
+                "split",
+                ("train", "valid", "test", "all"),
+            )
+        ]
         splits = self._load_csv("list_eval_partition.txt")
         identity = self._load_csv("identity_CelebA.txt")
         bbox = self._load_csv("list_bbox_celeba.txt", header=1)

torchvision/datasets/flowers102.py

@@ -112,3 +112,108 @@ def download(self):
         for id in ["label", "setid"]:
             filename, md5 = self._file_dict[id]
             download_url(self._download_url_prefix + filename, str(self._base_folder), md5=md5)
+
+    classes = [
+        "pink primrose",
+        "hard-leaved pocket orchid",
+        "canterbury bells",
+        "sweet pea",
+        "english marigold",
+        "tiger lily",
+        "moon orchid",
+        "bird of paradise",
+        "monkshood",
+        "globe thistle",
+        "snapdragon",
+        "colt's foot",
+        "king protea",
+        "spear thistle",
+        "yellow iris",
+        "globe-flower",
+        "purple coneflower",
+        "peruvian lily",
+        "balloon flower",
+        "giant white arum lily",
+        "fire lily",
+        "pincushion flower",
+        "fritillary",
+        "red ginger",
+        "grape hyacinth",
+        "corn poppy",
+        "prince of wales feathers",
+        "stemless gentian",
+        "artichoke",
+        "sweet william",
+        "carnation",
+        "garden phlox",
+        "love in the mist",
+        "mexican aster",
+        "alpine sea holly",
+        "ruby-lipped cattleya",
+        "cape flower",
+        "great masterwort",
+        "siam tulip",
+        "lenten rose",
+        "barbeton daisy",
+        "daffodil",
+        "sword lily",
+        "poinsettia",
+        "bolero deep blue",
+        "wallflower",
+        "marigold",
+        "buttercup",
+        "oxeye daisy",
+        "common dandelion",
+        "petunia",
+        "wild pansy",
+        "primula",
+        "sunflower",
+        "pelargonium",
+        "bishop of llandaff",
+        "gaura",
+        "geranium",
+        "orange dahlia",
+        "pink-yellow dahlia?",
+        "cautleya spicata",
+        "japanese anemone",
+        "black-eyed susan",
+        "silverbush",
+        "californian poppy",
+        "osteospermum",
+        "spring crocus",
+        "bearded iris",
+        "windflower",
+        "tree poppy",
+        "gazania",
+        "azalea",
+        "water lily",
+        "rose",
+        "thorn apple",
+        "morning glory",
+        "passion flower",
+        "lotus",
+        "toad lily",
+        "anthurium",
+        "frangipani",
+        "clematis",
+        "hibiscus",
+        "columbine",
+        "desert-rose",
+        "tree mallow",
+        "magnolia",
+        "cyclamen",
+        "watercress",
+        "canna lily",
+        "hippeastrum",
+        "bee balm",
+        "ball moss",
+        "foxglove",
+        "bougainvillea",
+        "camellia",
+        "mallow",
+        "mexican petunia",
+        "bromelia",
+        "blanket flower",
+        "trumpet creeper",
+        "blackberry lily",
+    ]

torchvision/datasets/mnist.py

@@ -25,18 +25,18 @@ class MNIST(VisionDataset):
             and  ``MNIST/raw/t10k-images-idx3-ubyte`` exist.
         train (bool, optional): If True, creates dataset from ``train-images-idx3-ubyte``,
             otherwise from ``t10k-images-idx3-ubyte``.
-        download (bool, optional): If True, downloads the dataset from the internet and
-            puts it in root directory. If dataset is already downloaded, it is not
-            downloaded again.
         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.
+        download (bool, optional): If True, downloads the dataset from the internet and
+            puts it in root directory. If dataset is already downloaded, it is not
+            downloaded again.
     """
 
     mirrors = [
-        "http://yann.lecun.com/exdb/mnist/",
         "https://ossci-datasets.s3.amazonaws.com/mnist/",
+        "http://yann.lecun.com/exdb/mnist/",
     ]
 
     resources = [
@@ -209,13 +209,13 @@ class FashionMNIST(MNIST):
             and  ``FashionMNIST/raw/t10k-images-idx3-ubyte`` exist.
         train (bool, optional): If True, creates dataset from ``train-images-idx3-ubyte``,
             otherwise from ``t10k-images-idx3-ubyte``.
-        download (bool, optional): If True, downloads the dataset from the internet and
-            puts it in root directory. If dataset is already downloaded, it is not
-            downloaded again.
         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.
+        download (bool, optional): If True, downloads the dataset from the internet and
+            puts it in root directory. If dataset is already downloaded, it is not
+            downloaded again.
     """
 
     mirrors = ["http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/"]
@@ -237,13 +237,13 @@ class KMNIST(MNIST):
             and  ``KMNIST/raw/t10k-images-idx3-ubyte`` exist.
         train (bool, optional): If True, creates dataset from ``train-images-idx3-ubyte``,
             otherwise from ``t10k-images-idx3-ubyte``.
-        download (bool, optional): If True, downloads the dataset from the internet and
-            puts it in root directory. If dataset is already downloaded, it is not
-            downloaded again.
         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.
+        download (bool, optional): If True, downloads the dataset from the internet and
+            puts it in root directory. If dataset is already downloaded, it is not
+            downloaded again.
     """
 
     mirrors = ["http://codh.rois.ac.jp/kmnist/dataset/kmnist/"]
@@ -358,6 +358,9 @@ class QMNIST(MNIST):
             for each example is class number (for compatibility with
             the MNIST dataloader) or a torch vector containing the
             full qmnist information. Default=True.
+        train (bool,optional,compatibility): When argument 'what' is
+            not specified, this boolean decides whether to load the
+            training set or the testing set.  Default: True.
         download (bool, optional): If True, downloads the dataset from
             the internet and puts it in root directory. If dataset is
             already downloaded, it is not downloaded again.
@@ -366,9 +369,6 @@ class QMNIST(MNIST):
             version. E.g, ``transforms.RandomCrop``
         target_transform (callable, optional): A function/transform
             that takes in the target and transforms it.
-        train (bool,optional,compatibility): When argument 'what' is
-            not specified, this boolean decides whether to load the
-            training set or the testing set.  Default: True.
     """
 
     subsets = {"train": "train", "test": "test", "test10k": "test", "test50k": "test", "nist": "nist"}
@@ -514,7 +514,7 @@ def read_sn3_pascalvincent_tensor(path: str, strict: bool = True) -> torch.Tenso
         data = f.read()
 
     # parse
-    if sys.byteorder == "little":
+    if sys.byteorder == "little" or sys.platform == "aix":
         magic = get_int(data[0:4])
         nd = magic % 256
         ty = magic // 256
@@ -527,7 +527,7 @@ def read_sn3_pascalvincent_tensor(path: str, strict: bool = True) -> torch.Tenso
     torch_type = SN3_PASCALVINCENT_TYPEMAP[ty]
     s = [get_int(data[4 * (i + 1) : 4 * (i + 2)]) for i in range(nd)]
 
-    if sys.byteorder == "big":
+    if sys.byteorder == "big" and not sys.platform == "aix":
         for i in range(len(s)):
             s[i] = int.from_bytes(s[i].to_bytes(4, byteorder="little"), byteorder="big", signed=False)
 

torchvision/datasets/moving_mnist.py

@@ -18,11 +18,11 @@ class MovingMNIST(VisionDataset):
         split_ratio (int, optional): The split ratio of number of frames. If ``split="train"``, the first split
             frames ``data[:, :split_ratio]`` is returned. If ``split="test"``, the last split frames ``data[:, split_ratio:]``
             is returned. If ``split=None``, this parameter is ignored and the all frames data is returned.
-        transform (callable, optional): A function/transform that takes in a torch Tensor
-            and returns a transformed version. E.g, ``transforms.RandomCrop``
         download (bool, optional): If true, downloads the dataset from the internet and
             puts it in root directory. If dataset is already downloaded, it is not
             downloaded again.
+        transform (callable, optional): A function/transform that takes in a torch Tensor
+            and returns a transformed version. E.g, ``transforms.RandomCrop``
     """
 
     _URL = "http://www.cs.toronto.edu/~nitish/unsupervised_video/mnist_test_seq.npy"

torchvision/datasets/oxford_iiit_pet.py

@@ -27,6 +27,8 @@ class OxfordIIITPet(VisionDataset):
         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.
+        transforms (callable, optional): A function/transform that takes input sample
+            and its target as entry and returns a transformed version.
         download (bool, optional): If True, downloads the dataset from the internet and puts it into
             ``root/oxford-iiit-pet``. If dataset is already downloaded, it is not downloaded again.
     """

torchvision/models/_api.py

@@ -7,7 +7,7 @@
 from functools import partial
 from inspect import signature
 from types import ModuleType
-from typing import Any, Callable, Dict, Iterable, List, Mapping, Optional, Set, Type, TypeVar, Union
+from typing import Any, Callable, Dict, get_args, Iterable, List, Mapping, Optional, Set, Type, TypeVar, Union
 
 from torch import nn
 
@@ -168,14 +168,13 @@ def _get_enum_from_fn(fn: Callable) -> Type[WeightsEnum]:
     if "weights" not in sig.parameters:
         raise ValueError("The method is missing the 'weights' argument.")
 
-    ann = signature(fn).parameters["weights"].annotation
+    ann = sig.parameters["weights"].annotation
     weights_enum = None
     if isinstance(ann, type) and issubclass(ann, WeightsEnum):
         weights_enum = ann
     else:
         # handle cases like Union[Optional, T]
-        # TODO: Replace ann.__args__ with typing.get_args(ann) after python >= 3.8
-        for t in ann.__args__:  # type: ignore[union-attr]
+        for t in get_args(ann):  # type: ignore[union-attr]
             if isinstance(t, type) and issubclass(t, WeightsEnum):
                 weights_enum = t
                 break

torchvision/ops/focal_loss.py

@@ -20,7 +20,7 @@ def sigmoid_focal_loss(
         targets (Tensor): A float tensor with the same shape as inputs. Stores the binary
                 classification label for each element in inputs
                 (0 for the negative class and 1 for the positive class).
-        alpha (float): Weighting factor in range (0,1) to balance
+        alpha (float): Weighting factor in range [0, 1] to balance
                 positive vs negative examples or -1 for ignore. Default: ``0.25``.
         gamma (float): Exponent of the modulating factor (1 - p_t) to
                 balance easy vs hard examples. Default: ``2``.
@@ -33,6 +33,9 @@ def sigmoid_focal_loss(
     """
     # Original implementation from https://github.com/facebookresearch/fvcore/blob/master/fvcore/nn/focal_loss.py
 
+    if not (0 <= alpha <= 1) or alpha != -1:
+        raise ValueError(f"Invalid alpha value: {alpha}. alpha must be in the range [0,1] or -1 for ignore.")
+
     if not torch.jit.is_scripting() and not torch.jit.is_tracing():
         _log_api_usage_once(sigmoid_focal_loss)
     p = torch.sigmoid(inputs)