* Upgrade presets for CUDA 12.6.0, cuDNN 9.3.0, NCCL 2.22.3, nvCOMP 4.0.0, NVIDIA Video Codec SDK 12.2.72, TensorRT 10.3.0.26

Author: saudet

.github/actions/deploy-ubuntu/action.yml

@@ -41,29 +41,29 @@ runs:
           export ARCH=arm64
           export ARCH_CUDA=sbsa
           export PREFIX=aarch64-linux-gnu
-          export CUDA=cuda-repo-rhel8-12-3-local-12.3.2_545.23.08-1.aarch64.rpm
-          export CUDNN=8.9.7.29-1.cuda12.2.aarch64
-          export NCCL=2.19.3-1+cuda12.3.aarch64
-          export NVCOMP=nvcomp_3.0.5_SBSA_12.x
+          export CUDA=cuda-repo-rhel8-12-6-local-12.6.0_560.28.03-1.aarch64.rpm
+          export CUDNN=cuda-12-9.3.0.75-1.aarch64
+          export NCCL=2.22.3-1+cuda12.5.aarch64
+          export NVCOMP=nvcomp-linux-sbsa-4.0.0-cuda12.5
           export USERLAND_BUILDME="buildme --aarch64"
         elif [[ "$CI_DEPLOY_PLATFORM" == "linux-ppc64le" ]]; then
           export ARCH=ppc64el
           export ARCH_CUDA=ppc64le
           export PREFIX=powerpc64le-linux-gnu
-          export CUDA=cuda-repo-rhel8-12-3-local-12.3.2_545.23.08-1.ppc64le.rpm
-          export CUDNN=8.9.7.29-1.cuda12.2.ppc64le
-          export NCCL=2.19.3-1+cuda12.3.ppc64le
+          export CUDA=cuda-repo-rhel8-12-6-local-12.6.0_560.28.03-1.ppc64le.rpm
+          export CUDNN=cuda-12-9.3.0.75-1.ppc64le
+          export NCCL=2.22.3-1+cuda12.5.ppc64le
         elif [[ "$CI_DEPLOY_PLATFORM" == "linux-x86" ]]; then
           export ARCH=i386
           export PREFIX=i686-linux-gnu
         elif [[ "$CI_DEPLOY_PLATFORM" == "linux-x86_64" ]]; then
           export ARCH=amd64
           export ARCH_CUDA=x86_64
           export PREFIX=x86_64-linux-gnu
-          export CUDA=cuda-repo-rhel8-12-3-local-12.3.2_545.23.08-1.x86_64.rpm
-          export CUDNN=8.9.7.29-1.cuda12.2.x86_64
-          export NCCL=2.19.3-1+cuda12.3.x86_64
-          export NVCOMP=nvcomp_3.0.5_x86_64_12.x
+          export CUDA=cuda-repo-rhel8-12-6-local-12.6.0_560.28.03-1.x86_64.rpm
+          export CUDNN=cuda-12-9.3.0.75-1.x86_64
+          export NCCL=2.22.3-1+cuda12.5.x86_64
+          export NVCOMP=nvcomp-linux-x86_64-4.0.0-cuda12.5
         fi
         echo "ARCH=$ARCH" >> $GITHUB_ENV
         echo "PREFIX=$PREFIX" >> $GITHUB_ENV
@@ -162,26 +162,26 @@ runs:
 
         if [[ -n ${ARCH_CUDA:-} ]] && [[ -n ${CI_DEPLOY_NEED_CUDA:-} ]]; then
           echo Installing CUDA, cuDNN, nvCOMP, etc
-          curl -LO https://developer.download.nvidia.com/compute/cuda/12.3.2/local_installers/$CUDA
-          curl -LO https://developer.download.nvidia.com/compute/cuda/repos/rhel8/$ARCH_CUDA/libcudnn8-$CUDNN.rpm
-          curl -LO https://developer.download.nvidia.com/compute/cuda/repos/rhel8/$ARCH_CUDA/libcudnn8-devel-$CUDNN.rpm
+          curl -LO https://developer.download.nvidia.com/compute/cuda/12.6.0/local_installers/$CUDA
+          curl -LO https://developer.download.nvidia.com/compute/cuda/repos/rhel8/$ARCH_CUDA/libcudnn9-$CUDNN.rpm
+          curl -LO https://developer.download.nvidia.com/compute/cuda/repos/rhel8/$ARCH_CUDA/libcudnn9-devel-$CUDNN.rpm
           curl -LO https://developer.download.nvidia.com/compute/cuda/repos/rhel8/$ARCH_CUDA/libnccl-$NCCL.rpm
           curl -LO https://developer.download.nvidia.com/compute/cuda/repos/rhel8/$ARCH_CUDA/libnccl-devel-$NCCL.rpm
 
           $SUDO rpm -i --force --ignorearch --nodeps $CUDA libcudnn*.rpm libnccl*.rpm
           rm -f *.rpm *.tgz *.txz *.tar.*
-          pushd /var/cuda-repo-rhel8-12-3-local/; $SUDO rpm -i --force --ignorearch --nodeps cuda*.rpm libc*.rpm libn*.rpm; $SUDO rm *.rpm; popd
+          pushd /var/cuda-repo-rhel8-12-6-local/; $SUDO rpm -i --force --ignorearch --nodeps cuda*.rpm libc*.rpm libn*.rpm; $SUDO rm *.rpm; popd
           $SUDO ln -sf /usr/local/cuda/lib64/ /usr/local/cuda/lib
           $SUDO ln -sf /usr/local/cuda/lib64/stubs/libcuda.so /usr/local/cuda/lib64/libcuda.so
           $SUDO ln -sf /usr/local/cuda/lib64/stubs/libnvidia-ml.so /usr/local/cuda/lib64/libnvidia-ml.so
           $SUDO mv /usr/include/cudnn* /usr/include/nccl* /usr/local/cuda/include/
           $SUDO mv /usr/lib64/libcudnn* /usr/lib64/libnccl* /usr/local/cuda/lib64/
 
           if [[ -n ${NVCOMP:-} ]]; then
-            curl -LO https://developer.download.nvidia.com/compute/nvcomp/3.0.5/local_installers/$NVCOMP.tgz
-            $SUDO tar -xvf $NVCOMP.tgz -C /usr/local/cuda/lib64/ --strip-components=1 lib/ || $SUDO tar -xvf $NVCOMP.tgz -C /usr/local/cuda/lib64/ --strip-components=2 nvcomp-3.0.5-ctk-12.2/lib/
-            $SUDO tar -xvf $NVCOMP.tgz -C /usr/local/cuda/include/ --strip-components=1 include/ || $SUDO tar -xvf $NVCOMP.tgz -C /usr/local/cuda/include/ --strip-components=2 nvcomp-3.0.5-ctk-12.2/include/
-            rm -f $NVCOMP.tgz
+            curl -LO https://developer.download.nvidia.com/compute/nvcomp/4.0.0/local_installers/$NVCOMP.tar.gz
+            $SUDO tar -xvf $NVCOMP.tar.gz -C /usr/local/cuda/lib64/ --strip-components=1 lib/ || $SUDO tar -xvf $NVCOMP.tar.gz -C /usr/local/cuda/lib64/ --strip-components=2 nvcomp/lib/
+            $SUDO tar -xvf $NVCOMP.tar.gz -C /usr/local/cuda/include/ --strip-components=1 include/ || $SUDO tar -xvf $NVCOMP.tar.gz -C /usr/local/cuda/include/ --strip-components=2 nvcomp/include/
+            rm -f $NVCOMP.tar.gz
           fi
 
           # Work around issues with CUDA 10.2/11.x
@@ -215,24 +215,24 @@ runs:
 
         if [[ "$CI_DEPLOY_MODULE" == "nvcodec" ]]; then
           echo Installing NVIDIA Video Codec SDK
-          python3 -m gdown 13KyYoq8ZOLIA7jugDkdFma34mnyxnz4x
-          $SUDO unzip Video_Codec_SDK_12.1.14.zip -d /usr/local
+          python3 -m gdown 15HMCd894BkX4N7e3RHKGpq0xVaEgxUB7
+          $SUDO unzip Video_Codec_SDK_12.2.72.zip-d /usr/local
           $SUDO ln -sf /usr/local/Video_Codec_SDK_* /usr/local/videocodecsdk
         fi
 
         if [[ "$CI_DEPLOY_PLATFORM" == "linux-arm64" ]] && [[ "$CI_DEPLOY_MODULE" == "tensorrt" ]]; then
           echo Installing TensorRT
           # python3 -m gdown 1LZRCv4ZAGiDQAu4pvADJIGntq4cGl5tU
-          curl -LO https://developer.download.nvidia.com/compute/machine-learning/tensorrt/10.0.1/tars/TensorRT-10.0.1.6.Ubuntu-22.04.aarch64-gnu.cuda-12.4.tar.gz
-          $SUDO tar -hxvf TensorRT-10.0.1.6.Ubuntu-22.04.aarch64-gnu.cuda-12.4.tar.gz -C /usr/local/
+          curl -LO https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.3.0/tars/TensorRT-10.3.0.26.Ubuntu-22.04.aarch64-gnu.cuda-12.5.tar.gz
+          $SUDO tar -hxvf TensorRT-10.3.0.26.Ubuntu-22.04.aarch64-gnu.cuda-12.5.tar.gz -C /usr/local/
           $SUDO ln -sf /usr/local/TensorRT* /usr/local/tensorrt
         fi
 
         if [[ "$CI_DEPLOY_PLATFORM" == "linux-x86_64" ]] && [[ "$CI_DEPLOY_MODULE" == "tensorrt" ]]; then
           echo Installing TensorRT
           # python3 -m gdown 1dVhD-DEYY42QbZe1GXl-vxe3k6KqWGsL
-          curl -LO https://developer.download.nvidia.com/compute/machine-learning/tensorrt/10.0.1/tars/TensorRT-10.0.1.6.Linux.x86_64-gnu.cuda-12.4.tar.gz
-          $SUDO tar -hxvf TensorRT-10.0.1.6.Linux.x86_64-gnu.cuda-12.4.tar.gz -C /usr/local/
+          curl -LO https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.3.0/tars/TensorRT-10.3.0.26.Linux.x86_64-gnu.cuda-12.5.tar.gz
+          $SUDO tar -hxvf TensorRT-10.3.0.26.Linux.x86_64-gnu.cuda-12.5.tar.gz -C /usr/local/
           $SUDO ln -sf /usr/local/TensorRT* /usr/local/tensorrt
         fi
 

.github/actions/deploy-windows/action.yml

@@ -99,60 +99,54 @@ runs:
         if "%CI_DEPLOY_PLATFORM%"=="windows-x86_64" if not "%CI_DEPLOY_NEED_CUDA%"=="" (
           echo Installing CUDA, cuDNN, nvCOMP, etc
           curl -LO https://developer.download.nvidia.com/compute/cuda/11.8.0/local_installers/cuda_11.8.0_522.06_windows.exe
-          curl -LO https://developer.download.nvidia.com/compute/cuda/12.3.2/local_installers/cuda_12.3.2_546.12_windows.exe
-          rem curl -LO https://developer.download.nvidia.com/compute/redist/cudnn/v8.8.0/local_installers/12.0/cudnn_8.8.0.121_windows.exe
-          rem python -m gdown 135Z7zfwguQe6vn7p013HtVkHFu9-_rru
-          curl -LO https://github.com/bytedeco/binaries/releases/download/1.5.10/cudnn-windows-x86_64-8.9.7.29_cuda12-archive.zip
+          curl -LO https://developer.download.nvidia.com/compute/cuda/12.6.0/local_installers/cuda_12.6.0_560.76_windows.exe
+          curl -LO https://developer.download.nvidia.com/compute/cudnn/redist/cudnn/windows-x86_64/cudnn-windows-x86_64-9.3.0.75_cuda12-archive.zip
           curl -LO http://www.winimage.com/zLibDll/zlib123dllx64.zip
-          curl -LO https://developer.download.nvidia.com/compute/nvcomp/3.0.5/local_installers/nvcomp_3.0.5_windows_12.x.zip
+          curl -LO https://developer.download.nvidia.com/compute/nvcomp/4.0.0/local_installers/nvcomp-windows-x86_64-4.0.0-cuda12.5.zip
           cuda_11.8.0_522.06_windows.exe -s
           bash -c "rm -Rf 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.8'"
           bash -c "mv 'C:/Program Files/NVIDIA Corporation/NvToolsExt' 'C:/Program Files/NVIDIA Corporation/NvToolsExt_old'"
-          cuda_12.3.2_546.12_windows.exe -s
+          cuda_12.6.0_560.76_windows.exe -s
           bash -c "mv 'C:/Program Files/NVIDIA Corporation/NvToolsExt_old' 'C:/Program Files/NVIDIA Corporation/NvToolsExt'"
           bash -c "ls 'C:/Program Files/NVIDIA Corporation/NvToolsExt'"
-          rem cudnn_8.8.0.121_windows.exe -s
-          unzip cudnn-windows-x86_64-8.9.7.29_cuda12-archive.zip
+          unzip cudnn-windows-x86_64-9.3.0.75_cuda12-archive.zip
           unzip zlib123dllx64.zip
-          unzip nvcomp_3.0.5_windows_12.x.zip
-          rem move "%ProgramFiles%\NVIDIA\CUDNN\v8.8\bin\*.dll" "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\bin"
-          rem move "%ProgramFiles%\NVIDIA\CUDNN\v8.8\include\*.h" "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\include"
-          rem move "%ProgramFiles%\NVIDIA\CUDNN\v8.8\lib\x64\*.lib" "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\lib\x64"
-          move cudnn-windows-x86_64-8.9.7.29_cuda12-archive\bin\*.dll "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\bin"
-          move cudnn-windows-x86_64-8.9.7.29_cuda12-archive\include\*.h "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\include"
-          move cudnn-windows-x86_64-8.9.7.29_cuda12-archive\lib\x64\*.lib "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\lib\x64"
-          move dll_x64\zlibwapi.dll "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\bin"
-          move include\* "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\include"
-          move include\gdeflate "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\include"
-          move include\native "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\include"
-          move include\nvcomp "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\include"
-          move lib\nvcomp*.dll "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\bin"
-          move lib\nvcomp*.lib "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\lib\x64"
+          unzip nvcomp-windows-x86_64-4.0.0-cuda12.5.zip
+          move cudnn-windows-x86_64-9.3.0.75_cuda12-archive\bin\*.dll "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6\bin"
+          move cudnn-windows-x86_64-9.3.0.75_cuda12-archive\include\*.h "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6\include"
+          move cudnn-windows-x86_64-9.3.0.75_cuda12-archive\lib\x64\*.lib "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6\lib\x64"
+          move dll_x64\zlibwapi.dll "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6\bin"
+          move nvcomp\include\* "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6\include"
+          move nvcomp\include\device "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6\include"
+          move nvcomp\include\native "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6\include"
+          move nvcomp\include\nvcomp "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6\include"
+          move nvcomp\lib\nvcomp*.dll "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6\bin"
+          move nvcomp\lib\nvcomp*.lib "%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6\lib\x64"
 
           rem echo Applying hotfix to Visual Studio 2019 for CUDA
           rem curl -LO https://raw.githubusercontent.com/microsoft/STL/main/stl/inc/cmath
           rem bash -c "find 'C:/Program Files (x86)/Microsoft Visual Studio/2019/Enterprise/VC/' -name cmath -exec cp -v cmath {} \;"
-          bash -c "sed -i 's/cublas_v2.h/cublas_api.h/g' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.3/include/cublasXt.h' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.3/include/cusolverDn.h' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.3/include/cusolverSp.h'"
-          bash -c "sed -i 's/#if !defined(__cplusplus) || __cplusplus >= 201103L/#if 1/g' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.3/include/driver_types.h'"
-          bash -c "sed -i '/block_merge_sort.cuh/d' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.3/include/cub/cub.cuh'"
-          bash -c "sed -i '/device_merge_sort.cuh/d' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.3/include/cub/cub.cuh'"
-          bash -c "sed -i '/device_segmented_sort.cuh/d' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.3/include/cub/cub.cuh'"
-          bash -c "sed -i '/warp_merge_sort.cuh/d' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.3/include/cub/cub.cuh'"
+          bash -c "sed -i 's/cublas_v2.h/cublas_api.h/g' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.6/include/cublasXt.h' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.6/include/cusolverDn.h' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.6/include/cusolverSp.h'"
+          bash -c "sed -i 's/#if !defined(__cplusplus) || __cplusplus >= 201103L/#if 1/g' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.6/include/driver_types.h'"
+          bash -c "sed -i '/block_merge_sort.cuh/d' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.6/include/cub/cub.cuh'"
+          bash -c "sed -i '/device_merge_sort.cuh/d' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.6/include/cub/cub.cuh'"
+          bash -c "sed -i '/device_segmented_sort.cuh/d' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.6/include/cub/cub.cuh'"
+          bash -c "sed -i '/warp_merge_sort.cuh/d' 'C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.6/include/cub/cub.cuh'"
         )
 
         if "%CI_DEPLOY_MODULE%"=="nvcodec" (
           echo Installing NVIDIA Video Codec SDK
-          python -m gdown 13KyYoq8ZOLIA7jugDkdFma34mnyxnz4x
-          unzip Video_Codec_SDK_12.1.14.zip
-          move Video_Codec_SDK_12.1.14 "%ProgramFiles%\NVIDIA GPU Computing Toolkit\VideoCodecSDK"
+          python -m gdown 15HMCd894BkX4N7e3RHKGpq0xVaEgxUB7
+          unzip Video_Codec_SDK_12.2.72.zip
+          move Video_Codec_SDK_12.2.72 "%ProgramFiles%\NVIDIA GPU Computing Toolkit\VideoCodecSDK"
         )
 
         if "%CI_DEPLOY_MODULE%"=="tensorrt" (
           echo Installing TensorRT
           rem python -m gdown 1GfmJ1BKbacLpUU-0i_mGu0sjrAS0Xzzi
-          curl -LO https://developer.download.nvidia.com/compute/machine-learning/tensorrt/10.0.1/zip/TensorRT-10.0.1.6.Windows10.win10.cuda-12.4.zip
-          unzip TensorRT-10.0.1.6.Windows10.win10.cuda-12.4.zip
-          move TensorRT-10.0.1.6 "%ProgramFiles%\NVIDIA GPU Computing Toolkit\TensorRT"
+          curl -LO https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/10.3.0/zip/TensorRT-10.3.0.26.Windows.win10.cuda-12.5.zip
+          unzip TensorRT-10.3.0.26.Windows.win10.cuda-12.5.zip
+          move TensorRT-10.3.0.26 "%ProgramFiles%\NVIDIA GPU Computing Toolkit\TensorRT"
         )
 
         if "%CI_DEPLOY_MODULE%"=="mkl" (
@@ -224,10 +218,10 @@ runs:
           C:/msys64/usr/bin/bazel.exe version
         )
         if exist "%ProgramFiles%\NVIDIA GPU Computing Toolkit" (
-          set "CUDA_PATH=%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3"
-          set "CUDA_PATH_V12_3=%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3"
-          set "PATH=%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\bin;%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\libnvvp;%PATH%"
-          echo CUDA Version 12.3.2>"%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.3\version.txt"
+          set "CUDA_PATH=%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6"
+          set "CUDA_PATH_V12_6=%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6"
+          set "PATH=%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6\bin;%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6\libnvvp;%PATH%"
+          echo CUDA Version 12.6.0>"%ProgramFiles%\NVIDIA GPU Computing Toolkit\CUDA\v12.6\version.txt"
         )
         set "CCACHE_DIR=%USERPROFILE%\ccache"
         set "PATH=C:\hostedtoolcache\windows\Python\3.9.13\x64;C:\msys64\%MSYSTEM%\bin;C:\msys64\usr\bin;%ProgramFiles%\apache-maven-3.6.3\bin;%PATH%"