lunar-walker edited a comment on issue #20256: URL: https://github.com/apache/incubator-mxnet/issues/20256#issuecomment-840005512
Build seems working fine with InsightFace pretrained models and performance is akin to running on Ubuntu... Is there any test suit available to thoroughly check a GPU build? And Should I close this issue now? Just for info for all others, here is the recap I managed to get Windows 10 x64 GPU Build of Version 1.8 by using VS 2017, CUDA 11.2 (Architecture 7.5), CUDNN, MKL, LAPACK & OpenCV by following:- 1. Installed all required frameworks & libraries & VS 2017 (Ensure there is CUDA & MSVC integration, hint: install CUDA after VS) 2. Use following command from x64 Native Command Prompt for Visual Studio 2017 'cmake -G "Visual Studio 15 2017 Win64" -T cuda=11.2,host=x64 -DCMAKE_BUILD_TYPE=Release -DUSE_CUDA=1 -DUSE_CUDNN=1 -DUSE_OPENCV=1 -DUSE_OPENMP=1 -DUSE_LAPACK=1 -DUSE_BLAS=mkl -DUSE_DIST_KVSTORE=0 -DCUDNN_INCLUDE=C:\cudnn\cuda\include -DCUDNN_LIBRARY=C:\cudnn\cuda\lib\x64\cudnn.lib -DMKL_ROOT="D:\oneAPI\mkl\latest" "d:\code\apache-mxnet-src-1.8.0-incubating" ' (Change appropriate paths as your configuration) 3. Use VS 2017 GUI to build mxnet.sln, after successful build, you ll find libraries in Build\Release folder 4. Copy all the dependencies DLLs as per your build parameters (i.e. CUDA, CUDNN, MKL, OpenCV, LAPACK etc) inside Release folder 5. Download and extract https://sourceforge.net/projects/openblas/files/v0.2.12/mingw64_dll.zip/download in the Release Folder. 5. Use Python\setup.py to install the Python bindings by invoking the script from inside python environment where you plan to install it. 5. Hope & Pray that it works Best Regards, -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@mxnet.apache.org For additional commands, e-mail: issues-h...@mxnet.apache.org
