Merge branch 'main' into patch-1

Author: ain-soph

.circleci/config.yml

@@ -32,8 +32,8 @@ def build_workflows(prefix="", filter_branch=None, upload=False, indentation=6,
         for os_type in ["linux", "macos", "win"]:
             python_versions = PYTHON_VERSIONS
             cu_versions_dict = {
-                "linux": ["cpu", "cu102", "cu111", "cu113", "rocm4.2", "rocm4.3.1"],
-                "win": ["cpu", "cu111", "cu113"],
+                "linux": ["cpu", "cu102", "cu111", "cu113", "cu115", "rocm4.2", "rocm4.3.1"],
+                "win": ["cpu", "cu111", "cu113", "cu115"],
                 "macos": ["cpu"],
             }
             cu_versions = cu_versions_dict[os_type]
@@ -128,6 +128,7 @@ def upload_doc_job(filter_branch):
     "cu111": "pytorch/manylinux-cuda111",
     "cu112": "pytorch/manylinux-cuda112",
     "cu113": "pytorch/manylinux-cuda113",
+    "cu115": "pytorch/manylinux-cuda115",
 }
 
 

.circleci/regenerate.py

@@ -11,5 +11,11 @@ setup_conda_pytorch_constraint
 setup_conda_cudatoolkit_constraint
 setup_visual_studio_constraint
 setup_junit_results_folder
-# nvidia channel included for cudatoolkit >= 11
-conda build -c defaults -c nvidia $CONDA_CHANNEL_FLAGS --no-anaconda-upload --python "$PYTHON_VERSION" packaging/torchvision
+
+# nvidia channel included for cudatoolkit >= 11 however for 11.5 we use conda-forge
+export CUDATOOLKIT_CHANNEL="nvidia"
+if [[ "$CU_VERSION" == cu115 ]]; then
+    export CUDATOOLKIT_CHANNEL="conda-forge"
+fi
+
+conda build -c defaults -c $CUDATOOLKIT_CHANNEL $CONDA_CHANNEL_FLAGS --no-anaconda-upload --python "$PYTHON_VERSION" packaging/torchvision

packaging/build_conda.sh

@@ -46,6 +46,14 @@ setup_cuda() {
 
   # Now work out the CUDA settings
   case "$CU_VERSION" in
+    cu115)
+      if [[ "$OSTYPE" == "msys" ]]; then
+        export CUDA_HOME="C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v11.5"
+      else
+        export CUDA_HOME=/usr/local/cuda-11.5/
+      fi
+      export TORCH_CUDA_ARCH_LIST="3.5;5.0+PTX;6.0;7.0;7.5;8.0;8.6"
+      ;;
     cu113)
       if [[ "$OSTYPE" == "msys" ]]; then
         export CUDA_HOME="C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v11.3"
@@ -289,6 +297,9 @@ setup_conda_cudatoolkit_constraint() {
     export CONDA_BUILD_VARIANT="cpu"
   else
     case "$CU_VERSION" in
+      cu115)
+        export CONDA_CUDATOOLKIT_CONSTRAINT="- cudatoolkit >=11.5,<11.6 # [not osx]"
+        ;;
       cu113)
         export CONDA_CUDATOOLKIT_CONSTRAINT="- cudatoolkit >=11.3,<11.4 # [not osx]"
         ;;
@@ -333,6 +344,9 @@ setup_conda_cudatoolkit_plain_constraint() {
     export CMAKE_USE_CUDA=0
   else
     case "$CU_VERSION" in
+      cu115)
+        export CONDA_CUDATOOLKIT_CONSTRAINT="cudatoolkit=11.5"
+        ;;
       cu113)
         export CONDA_CUDATOOLKIT_CONSTRAINT="cudatoolkit=11.3"
         ;;

packaging/pkg_helpers.bash

@@ -21,6 +21,7 @@ set CUDA_VER_MAJOR=%CUDA_VER:~0,-1%
 set CUDA_VER_MINOR=%CUDA_VER:~-1,1%
 set CUDA_VERSION_STR=%CUDA_VER_MAJOR%.%CUDA_VER_MINOR%
 
+
 if %CUDA_VER% EQU 92 goto cuda92
 if %CUDA_VER% EQU 100 goto cuda100
 if %CUDA_VER% EQU 101 goto cuda101
@@ -29,6 +30,8 @@ if %CUDA_VER% EQU 110 goto cuda110
 if %CUDA_VER% EQU 111 goto cuda111
 if %CUDA_VER% EQU 112 goto cuda112
 if %CUDA_VER% EQU 113 goto cuda113
+if %CUDA_VER% EQU 115 goto cuda115
+
 
 echo CUDA %CUDA_VERSION_STR% is not supported
 exit /b 1
@@ -180,6 +183,25 @@ if not exist "%SRC_DIR%\temp_build\%CUDNN_INSTALL_ZIP%" (
 
 goto cuda_common
 
+:cuda115
+
+set CUDA_INSTALL_EXE=cuda_11.5.0_496.13_win10.exe
+if not exist "%SRC_DIR%\temp_build\%CUDA_INSTALL_EXE%" (
+    curl -k -L "https://ossci-windows.s3.amazonaws.com/%CUDA_INSTALL_EXE%" --output "%SRC_DIR%\temp_build\%CUDA_INSTALL_EXE%"
+    if errorlevel 1 exit /b 1
+    set "CUDA_SETUP_FILE=%SRC_DIR%\temp_build\%CUDA_INSTALL_EXE%"
+    set "ARGS=thrust_11.5 nvcc_11.5 cuobjdump_11.5 nvprune_11.5 nvprof_11.5 cupti_11.5 cublas_11.5 cublas_dev_11.5 cudart_11.5 cufft_11.5 cufft_dev_11.5 curand_11.5 curand_dev_11.5 cusolver_11.5 cusolver_dev_11.5 cusparse_11.5 cusparse_dev_11.5 npp_11.5 npp_dev_11.5 nvrtc_11.5 nvrtc_dev_11.5 nvml_dev_11.5"
+)
+
+set CUDNN_INSTALL_ZIP=cudnn-11.3-windows-x64-v8.2.0.53.zip
+if not exist "%SRC_DIR%\temp_build\%CUDNN_INSTALL_ZIP%" (
+    curl -k -L "http://s3.amazonaws.com/ossci-windows/%CUDNN_INSTALL_ZIP%" --output "%SRC_DIR%\temp_build\%CUDNN_INSTALL_ZIP%"
+    if errorlevel 1 exit /b 1
+    set "CUDNN_SETUP_FILE=%SRC_DIR%\temp_build\%CUDNN_INSTALL_ZIP%"
+)
+
+goto cuda_common
+
 :cuda_common
 
 if not exist "%SRC_DIR%\temp_build\NvToolsExt.7z" (
@@ -222,7 +244,7 @@ set "NVTOOLSEXT_PATH=%ProgramFiles%\NVIDIA Corporation\NvToolsExt\bin\x64"
 if not exist "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v%CUDA_VERSION_STR%\bin\nvcc.exe" (
     echo CUDA %CUDA_VERSION_STR% installed failed.
     echo --------- RunDll32.exe.log
-    type "%SRC_DIR%\temp_build\cuda\cuda_install_logs\LOG.RunDll32.exe.log"    
+    type "%SRC_DIR%\temp_build\cuda\cuda_install_logs\LOG.RunDll32.exe.log"
     echo --------- setup.exe.log -------
     type "%SRC_DIR%\temp_build\cuda\cuda_install_logs\LOG.setup.exe.log"
     exit /b 1

packaging/windows/internal/cuda_install.bat

@@ -35,7 +35,7 @@ def __init__(
         transform: Optional[Callable] = None,
         target_transform: Optional[Callable] = None,
     ) -> None:
-        _log_api_usage_once(self)
+        _log_api_usage_once("datasets", self.__class__.__name__)
         if isinstance(root, torch._six.string_classes):
             root = os.path.expanduser(root)
         self.root = root

torchvision/datasets/vision.py

@@ -18,7 +18,7 @@
 class AlexNet(nn.Module):
     def __init__(self, num_classes: int = 1000, dropout: float = 0.5) -> None:
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
         self.features = nn.Sequential(
             nn.Conv2d(3, 64, kernel_size=11, stride=4, padding=2),
             nn.ReLU(inplace=True),

torchvision/models/alexnet.py

@@ -163,7 +163,7 @@ def __init__(
     ) -> None:
 
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
 
         # First convolution
         self.features = nn.Sequential(

torchvision/models/densenet.py

@@ -27,7 +27,7 @@ class GeneralizedRCNN(nn.Module):
 
     def __init__(self, backbone: nn.Module, rpn: nn.Module, roi_heads: nn.Module, transform: nn.Module) -> None:
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
         self.transform = transform
         self.backbone = backbone
         self.rpn = rpn

torchvision/models/detection/generalized_rcnn.py

@@ -337,7 +337,7 @@ def __init__(
         topk_candidates=1000,
     ):
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
 
         if not hasattr(backbone, "out_channels"):
             raise ValueError(

torchvision/models/detection/retinanet.py

@@ -182,7 +182,7 @@ def __init__(
         positive_fraction: float = 0.25,
     ):
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
 
         self.backbone = backbone
 

torchvision/models/detection/ssd.py

@@ -120,7 +120,7 @@ def __init__(
         min_depth: int = 16,
     ):
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
 
         assert not backbone[c4_pos].use_res_connect
         self.features = nn.Sequential(

torchvision/models/detection/ssdlite.py

@@ -170,7 +170,7 @@ def __init__(
             norm_layer (Optional[Callable[..., nn.Module]]): Module specifying the normalization layer to use
         """
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
 
         if not inverted_residual_setting:
             raise ValueError("The inverted_residual_setting should not be empty")

torchvision/models/efficientnet.py

@@ -39,7 +39,7 @@ def __init__(
         dropout_aux: float = 0.7,
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
         if blocks is None:
             blocks = [BasicConv2d, Inception, InceptionAux]
         if init_weights is None:

torchvision/models/googlenet.py

@@ -37,7 +37,7 @@ def __init__(
         dropout: float = 0.5,
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
         if inception_blocks is None:
             inception_blocks = [BasicConv2d, InceptionA, InceptionB, InceptionC, InceptionD, InceptionE, InceptionAux]
         if init_weights is None:

torchvision/models/inception.py

@@ -98,7 +98,7 @@ class MNASNet(torch.nn.Module):
 
     def __init__(self, alpha: float, num_classes: int = 1000, dropout: float = 0.2) -> None:
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
         assert alpha > 0.0
         self.alpha = alpha
         self.num_classes = num_classes

torchvision/models/mnasnet.py

@@ -111,7 +111,7 @@ def __init__(
 
         """
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
 
         if block is None:
             block = InvertedResidual

torchvision/models/mobilenetv2.py

@@ -151,7 +151,7 @@ def __init__(
             dropout (float): The droupout probability
         """
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
 
         if not inverted_residual_setting:
             raise ValueError("The inverted_residual_setting should not be empty")

torchvision/models/mobilenetv3.py

@@ -440,7 +440,7 @@ def __init__(self, *, feature_encoder, context_encoder, corr_block, update_block
                 If ``None`` (default), the flow is upsampled using interpolation.
         """
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
 
         self.feature_encoder = feature_encoder
         self.context_encoder = context_encoder

torchvision/models/optical_flow/raft.py

@@ -310,7 +310,7 @@ def __init__(
         activation: Optional[Callable[..., nn.Module]] = None,
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
 
         if stem_type is None:
             stem_type = SimpleStemIN

torchvision/models/regnet.py

@@ -174,7 +174,7 @@ def __init__(
         norm_layer: Optional[Callable[..., nn.Module]] = None,
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
         if norm_layer is None:
             norm_layer = nn.BatchNorm2d
         self._norm_layer = norm_layer

torchvision/models/resnet.py

@@ -13,7 +13,7 @@ class _SimpleSegmentationModel(nn.Module):
 
     def __init__(self, backbone: nn.Module, classifier: nn.Module, aux_classifier: Optional[nn.Module] = None) -> None:
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
         self.backbone = backbone
         self.classifier = classifier
         self.aux_classifier = aux_classifier

torchvision/models/segmentation/_utils.py

@@ -38,7 +38,7 @@ def __init__(
         self, backbone: nn.Module, low_channels: int, high_channels: int, num_classes: int, inter_channels: int = 128
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
         self.backbone = backbone
         self.classifier = LRASPPHead(low_channels, high_channels, num_classes, inter_channels)
 

torchvision/models/segmentation/lraspp.py

@@ -100,7 +100,7 @@ def __init__(
         inverted_residual: Callable[..., nn.Module] = InvertedResidual,
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
 
         if len(stages_repeats) != 3:
             raise ValueError("expected stages_repeats as list of 3 positive ints")

torchvision/models/shufflenetv2.py

@@ -36,7 +36,7 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
 class SqueezeNet(nn.Module):
     def __init__(self, version: str = "1_0", num_classes: int = 1000, dropout: float = 0.5) -> None:
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
         self.num_classes = num_classes
         if version == "1_0":
             self.features = nn.Sequential(

torchvision/models/squeezenet.py

@@ -37,7 +37,7 @@ def __init__(
         self, features: nn.Module, num_classes: int = 1000, init_weights: bool = True, dropout: float = 0.5
     ) -> None:
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
         self.features = features
         self.avgpool = nn.AdaptiveAvgPool2d((7, 7))
         self.classifier = nn.Sequential(

torchvision/models/vgg.py

@@ -209,7 +209,7 @@ def __init__(
             zero_init_residual (bool, optional): Zero init bottleneck residual BN. Defaults to False.
         """
         super().__init__()
-        _log_api_usage_once(self)
+        _log_api_usage_once("models", self.__class__.__name__)
         self.inplanes = 64
 
         self.stem = stem()

torchvision/models/video/resnet.py

@@ -34,7 +34,7 @@ def nms(boxes: Tensor, scores: Tensor, iou_threshold: float) -> Tensor:
         Tensor: int64 tensor with the indices of the elements that have been kept
         by NMS, sorted in decreasing order of scores
     """
-    _log_api_usage_once("torchvision.ops.nms")
+    _log_api_usage_once("ops", "nms")
     _assert_has_ops()
     return torch.ops.torchvision.nms(boxes, scores, iou_threshold)
 
@@ -63,7 +63,7 @@ def batched_nms(
         Tensor: int64 tensor with the indices of the elements that have been kept by NMS, sorted
         in decreasing order of scores
     """
-    _log_api_usage_once("torchvision.ops.batched_nms")
+    _log_api_usage_once("ops", "batched_nms")
     # Benchmarks that drove the following thresholds are at
     # https://github.com/pytorch/vision/issues/1311#issuecomment-781329339
     if boxes.numel() > (4000 if boxes.device.type == "cpu" else 20000) and not torchvision._is_tracing():
@@ -122,7 +122,7 @@ def remove_small_boxes(boxes: Tensor, min_size: float) -> Tensor:
         Tensor[K]: indices of the boxes that have both sides
         larger than min_size
     """
-    _log_api_usage_once("torchvision.ops.remove_small_boxes")
+    _log_api_usage_once("ops", "remove_small_boxes")
     ws, hs = boxes[:, 2] - boxes[:, 0], boxes[:, 3] - boxes[:, 1]
     keep = (ws >= min_size) & (hs >= min_size)
     keep = torch.where(keep)[0]
@@ -141,7 +141,7 @@ def clip_boxes_to_image(boxes: Tensor, size: Tuple[int, int]) -> Tensor:
     Returns:
         Tensor[N, 4]: clipped boxes
     """
-    _log_api_usage_once("torchvision.ops.clip_boxes_to_image")
+    _log_api_usage_once("ops", "clip_boxes_to_image")
     dim = boxes.dim()
     boxes_x = boxes[..., 0::2]
     boxes_y = boxes[..., 1::2]
@@ -182,7 +182,7 @@ def box_convert(boxes: Tensor, in_fmt: str, out_fmt: str) -> Tensor:
         Tensor[N, 4]: Boxes into converted format.
     """
 
-    _log_api_usage_once("torchvision.ops.box_convert")
+    _log_api_usage_once("ops", "box_convert")
     allowed_fmts = ("xyxy", "xywh", "cxcywh")
     if in_fmt not in allowed_fmts or out_fmt not in allowed_fmts:
         raise ValueError("Unsupported Bounding Box Conversions for given in_fmt and out_fmt")
@@ -232,7 +232,7 @@ def box_area(boxes: Tensor) -> Tensor:
     Returns:
         Tensor[N]: the area for each box
     """
-    _log_api_usage_once("torchvision.ops.box_area")
+    _log_api_usage_once("ops", "box_area")
     boxes = _upcast(boxes)
     return (boxes[:, 2] - boxes[:, 0]) * (boxes[:, 3] - boxes[:, 1])
 
@@ -268,7 +268,7 @@ def box_iou(boxes1: Tensor, boxes2: Tensor) -> Tensor:
     Returns:
         Tensor[N, M]: the NxM matrix containing the pairwise IoU values for every element in boxes1 and boxes2
     """
-    _log_api_usage_once("torchvision.ops.box_iou")
+    _log_api_usage_once("ops", "box_iou")
     inter, union = _box_inter_union(boxes1, boxes2)
     iou = inter / union
     return iou
@@ -291,7 +291,7 @@ def generalized_box_iou(boxes1: Tensor, boxes2: Tensor) -> Tensor:
         for every element in boxes1 and boxes2
     """
 
-    _log_api_usage_once("torchvision.ops.generalized_box_iou")
+    _log_api_usage_once("ops", "generalized_box_iou")
     # degenerate boxes gives inf / nan results
     # so do an early check
     assert (boxes1[:, 2:] >= boxes1[:, :2]).all()
@@ -323,7 +323,7 @@ def masks_to_boxes(masks: torch.Tensor) -> torch.Tensor:
     Returns:
         Tensor[N, 4]: bounding boxes
     """
-    _log_api_usage_once("torchvision.ops.masks_to_boxes")
+    _log_api_usage_once("ops", "masks_to_boxes")
     if masks.numel() == 0:
         return torch.zeros((0, 4), device=masks.device, dtype=torch.float)