setUpNet DNN module was not built with CUDA backend; switching to CPU

REVRBE commented 1 week ago

Describe the bug Whenever I try to load a model using CUDA as target and backend it throws this runtime error stated in the title.

net_.setPreferableBackend(cv::dnn::DNN_BACKEND_CUDA); net_.setPreferableTarget(cv::dnn::DNN_TARGET_CUDA);

Environment

OS: Windows 10 x64]
CUDA 12.6
cuDNN 9.5
OpenCV 4.8.0

To Reproduce Steps to reproduce the behavior:

.\vcpkg.exe install opencv[cuda,contrib,cudnn,dnn,freetype,jpeg,openmp,png,webp,world]:x64-windows-static
No errors in building, error occurs at runtime

Expected behavior I expected it to initialize CUDA when loading the weights and config e.g. use GPU for faster inference . I'd expect the DNN module to have been built with CUDA given that I have used the proper modules in the vcpkg install command.

Additional context Fetched and updated vcpkg prior to installation.

I'm using it in Visual Studio Community 2022 in a C++ project, and have linked the following .lib files: cudart_static.lib (from CUDA directory), opencv_world4.lib (from vcpkg) and cudnn.lib, yet no success.

I assume this is a vcpkg issue, as I think I have installed the proper modules for building the DNN module with CUDA, but apparently not. I have also tried the module 'dnn-cuda' (it also is unable to install, trying to reproduce it right now, will update if it successfully builds or not: .\vcpkg.exe install opencv[dnn-cuda]:x64-windows-static --recurse) but I can't seem to find any useful information online on this particular module, most places they just say to use 'cuda', 'cudnn' and 'dnn' module in the vcpkg install command.

EDIT 1 https://imgur.com/a/10RAQYy So I tried to install the 'dnn-cuda' module only, like this: .\vcpkg.exe install opencv[dnn-cuda]:x64-windows-static --recurse but it just fails with the error on the image. If log files are needed, please help me and tell me which one specifically :)

EDIT 2 Output from: void print_opencv_info() { std::cout << "OpenCV version: " << cv::getBuildInformation() << std::endl; std::cout << "CUDA-enabled GPU devices: " << cv::cuda::getCudaEnabledDeviceCount() << std::endl; std::cout << "OpenCV version: " << CV_VERSION << std::endl; }

`OpenCV version: General configuration for OpenCV 4.8.0 ===================================== Version control: unknown

Extra modules: Location (extra): C:/vcpkg/vcpkg/buildtrees/opencv4/src/4.8.0-59c03c175c.clean/modules Version control (extra): unknown

Platform: Timestamp: 2024-10-18T08:47:03Z Host: Windows 10.0.19045 AMD64 CMake: 3.30.5 CMake generator: Ninja CMake build tool: C:/Program Files/Microsoft Visual Studio/2022/Community/Common7/IDE/CommonExtensions/Microsoft/CMake/Ninja/ninja.exe MSVC: 1937 Configuration: Release

CPU/HW features: Baseline: SSE SSE2 SSE3 requested: SSE3 Dispatched code generation: SSE4_1 SSE4_2 FP16 AVX AVX2 AVX512_SKX requested: SSE4_1 SSE4_2 AVX FP16 AVX2 AVX512_SKX SSE4_1 (14 files): + SSSE3 SSE4_1 SSE4_2 (1 files): + SSSE3 SSE4_1 POPCNT SSE4_2 FP16 (0 files): + SSSE3 SSE4_1 POPCNT SSE4_2 FP16 AVX AVX (7 files): + SSSE3 SSE4_1 POPCNT SSE4_2 AVX AVX2 (33 files): + SSSE3 SSE4_1 POPCNT SSE4_2 FP16 FMA3 AVX AVX2 AVX512_SKX (5 files): + SSSE3 SSE4_1 POPCNT SSE4_2 FP16 FMA3 AVX AVX2 AVX_512F AVX512_COMMON AVX512_SKX

C/C++: Built as dynamic libs?: NO C++ standard: 11 C++ Compiler: C:/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools/MSVC/14.37.32822/bin/Hostx64/x64/cl.exe (ver 19.37.32824.0) C++ flags (Release): /nologo /DWIN32 /D_WINDOWS /utf-8 /GR /MP /D _CRT_SECURE_NO_DEPRECATE /D _CRT_NONSTDC_NO_DEPRECATE /D _SCL_SECURE_NO_WARNINGS /Gy /bigobj /Oi /fp:precise /FS /EHa /wd4127 /wd4251 /wd4324 /wd4275 /wd4512 /wd4589 /wd4819 -openmp /MT /O2 /Oi /Gy /DNDEBUG /Z7 C++ flags (Debug): /nologo /DWIN32 /D_WINDOWS /utf-8 /GR /MP /D _CRT_SECURE_NO_DEPRECATE /D _CRT_NONSTDC_NO_DEPRECATE /D _SCL_SECURE_NO_WARNINGS /Gy /bigobj /Oi /fp:precise /FS /EHa /wd4127 /wd4251 /wd4324 /wd4275 /wd4512 /wd4589 /wd4819 -openmp /D_DEBUG /MTd /Z7 /Ob0 /Od /RTC1 C Compiler: C:/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools/MSVC/14.37.32822/bin/Hostx64/x64/cl.exe C flags (Release): /nologo /DWIN32 /D_WINDOWS /utf-8 /MP /D _CRT_SECURE_NO_DEPRECATE /D _CRT_NONSTDC_NO_DEPRECATE /D _SCL_SECURE_NO_WARNINGS /Gy /bigobj /Oi /fp:precise /FS -openmp /MT /O2 /Oi /Gy /DNDEBUG /Z7 C flags (Debug): /nologo /DWIN32 /D_WINDOWS /utf-8 /MP /D _CRT_SECURE_NO_DEPRECATE /D _CRT_NONSTDC_NO_DEPRECATE /D _SCL_SECURE_NO_WARNINGS /Gy /bigobj /Oi /fp:precise /FS -openmp /D_DEBUG /MTd /Z7 /Ob0 /Od /RTC1 Linker flags (Release): /machine:x64 /NODEFAULTLIB:atlthunk.lib /nologo /DEBUG /INCREMENTAL:NO /OPT:REF /OPT:ICF /debug /NODEFAULTLIB:libcmtd.lib /NODEFAULTLIB:libcpmtd.lib /NODEFAULTLIB:msvcrtd.lib Linker flags (Debug): /machine:x64 /NODEFAULTLIB:atlthunk.lib /nologo /debug /INCREMENTAL /NODEFAULTLIB:libcmt.lib /NODEFAULTLIB:libcpmt.lib /NODEFAULTLIB:msvcrt.lib ccache: NO Precompiled headers: NO Extra dependencies: protobuf::libprotobuf comctl32 gdi32 ole32 setupapi ws2_32 hdf5::hdf5-static harfbuzz::harfbuzz quirc::quirc cudart_static.lib nppc.lib nppial.lib nppicc.lib nppidei.lib nppif.lib nppig.lib nppim.lib nppist.lib nppisu.lib nppitc.lib npps.lib cublas.lib cudnn.lib cufft.lib -LIBPATH:"C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.6/lib/x64" 3rdparty dependencies:

OpenCV modules: To be built: aruco bgsegm bioinspired calib3d ccalib core cudaarithm cudabgsegm cudacodec cudafeatures2d cudafilters cudaimgproc cudalegacy cudaobjdetect cudaoptflow cudastereo cudawarping cudev datasets dnn dnn_objdetect dnn_superres dpm face features2d flann freetype fuzzy hdf hfs highgui img_hash imgcodecs imgproc intensity_transform line_descriptor mcc ml objdetect optflow phase_unwrapping photo plot quality rapid reg saliency shape stereo stitching structured_light superres surface_matching text tracking video videoio videostab wechat_qrcode world xfeatures2d ximgproc xobjdetect xphoto Disabled: rgbd Disabled by dependency: - Unavailable: alphamat cvv gapi java julia matlab ovis python2 python2 python3 sfm ts viz Applications: - Documentation: NO Non-free algorithms: NO

Windows RT support: NO

GUI: Win32 UI: YES

Media I/O: ZLib: optimized C:/vcpkg/vcpkg/installed/x64-windows-static/lib/zlib.lib debug C:/vcpkg/vcpkg/installed/x64-windows-static/debug/lib/zlibd.lib (ver 1.3.1) JPEG: optimized C:/vcpkg/vcpkg/installed/x64-windows-static/lib/jpeg.lib debug C:/vcpkg/vcpkg/installed/x64-windows-static/debug/lib/jpeg.lib (ver 62) WEBP: (ver 1.4.0) PNG: optimized C:/vcpkg/vcpkg/installed/x64-windows-static/lib/libpng16.lib debug C:/vcpkg/vcpkg/installed/x64-windows-static/debug/lib/libpng16d.lib (ver 1.6.43) TIFF: optimized C:/vcpkg/vcpkg/installed/x64-windows-static/lib/tiff.lib debug C:/vcpkg/vcpkg/installed/x64-windows-static/debug/lib/tiffd.lib (ver 42 / 4.7.0) HDR: YES SUNRASTER: YES PXM: YES PFM: YES

Video I/O: DirectShow: YES Media Foundation: YES DXVA: YES

Parallel framework: OpenMP

Trace: YES (built-in)

Other third-party libraries: Custom HAL: NO Protobuf: optimized C:/vcpkg/vcpkg/installed/x64-windows-static/lib/libprotobuf.lib debug C:/vcpkg/vcpkg/installed/x64-windows-static/debug/lib/libprotobufd.lib version (25.1.0) Flatbuffers: 23.5.26

NVIDIA CUDA: YES (ver 12.6, CUFFT CUBLAS) NVIDIA GPU arch: 50 52 60 61 70 75 80 86 89 90 NVIDIA PTX archs:

cuDNN: YES (ver )

OpenCL: YES (NVD3D11) Include path: C:/vcpkg/vcpkg/buildtrees/opencv4/src/4.8.0-f4e8005717.clean/3rdparty/include/opencl/1.2 Link libraries: Dynamic load

Python (for build): NO

Install to: C:/vcpkg/vcpkg/packages/opencv4_x64-windows-static

CUDA-enabled GPU devices: 1 OpenCV version: 4.8.0 [ WARN:0@0.079] global net_impl.cpp:178 cv::dnn::dnn4_v20230620::Net::Impl::setUpNet DNN module was not built with CUDA backend; switching to CPU`

WangWeiLin-MV commented 2 days ago

Environment

OS: Windows 10 x64]

CUDA 12.6

cuDNN 9.5

OpenCV 4.8.0

Build failed with cudnn/recurrent.hpp error ‘cudnnSetRNNDescriptor_v6’ was not declared, and this error doesn't seem to exist before cuDNN 9

Backport compat patch to build it.

diff --git a/ports/opencv4/portfile.cmake b/ports/opencv4/portfile.cmake
index 3f1de208d..5b89cc3c2 100644
--- a/ports/opencv4/portfile.cmake
+++ b/ports/opencv4/portfile.cmake
@@ -13,6 +13,12 @@ vcpkg_download_distfile(CUDA_12_4_FIX
   FILENAME opencv4-3e3ee106fb8ccd003aa2c9a943a2340b066537bc.patch
 )

+vcpkg_download_distfile(PATCH_PR_25412_COMPAT_CUDNN_9
+  URLS https://github.com/opencv/opencv/commit/05e48605a0aea00d3a89b9ab5e25cdf89568aa28.patch?full_index=1
+  SHA512 1229c8c9845302942884fac16b36c43dfdeca60c250bd4684ee31bff0346048ecd52f10f770dd033ae12bd578a6e165e71e6ebdd306e3d4d48099267bfefd91d
+  FILENAME opencv4-05e48605a0aea00d3a89b9ab5e25cdf89568aa28.patch
+)
+
 vcpkg_from_github(
     OUT_SOURCE_PATH SOURCE_PATH
     REPO opencv/opencv
@@ -39,6 +45,7 @@ vcpkg_from_github(
       "${ARM64_WINDOWS_FIX}"
       0022-fix-supportqnx.patch
       "${CUDA_12_4_FIX}"
+      "PATCH_PR_25412_COMPAT_CUDNN_9"
 )
 # Disallow accidental build of vendored copies
 file(REMOVE_RECURSE "${SOURCE_PATH}/3rdparty/openexr")

Build passed opencv[dnn-cuda]:x64-windows and `opencv[dnn-cuda]:x64-windows-static

MSVC: 19.41
CUDA: 12.5.0
cuDNN: 9.2.0

WangWeiLin-MV commented 2 days ago

@REVRBE Could you please try the patch above? We have no CUDA device for further testing.

microsoft / vcpkg

setUpNet DNN module was not built with CUDA backend; switching to CPU #41642

Install to: C:/vcpkg/vcpkg/packages/opencv4_x64-windows-static