[NVIDIA] CUDA version update (openvinotoolkit#652)

* update the dockerfile to use cuda base image and correct apt nvidia packages * Return cudnn into Dockerfile, adds drivers * Update README.md --------- Co-authored-by: Dariusz Trawinski <[email protected]>
sunxiaoxia2022 · May 8, 2023 · 032afce · 032afce
1 parent 4c116a1
commit 032afce
Show file tree

Hide file tree

Showing 2 changed files with 17 additions and 14 deletions.
diff --git a/modules/nvidia_plugin/Dockerfile b/modules/nvidia_plugin/Dockerfile
@@ -1,7 +1,8 @@
-FROM docker.io/library/ubuntu:focal
+FROM nvidia/cuda:11.8.0-runtime-ubuntu20.04
 ARG enable_tensorrt
 ADD https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600
 
+# nv-tensorrt-repo-*.deb packages
 COPY *.deb  /
 
 SHELL ["/bin/bash", "-c"]
@@ -24,7 +25,8 @@ RUN set -exuo pipefail ; \
 		build-essential \
 		ca-certificates \
 		clang-format-9 \
-		cuda \
+		cuda-runtime-11-8 \
+		cuda-11-8 \
 		curl \
 		g++-multilib \
 		gcc-multilib \
@@ -35,8 +37,8 @@ RUN set -exuo pipefail ; \
 		libavformat-dev \
 		libboost-regex-dev \
 		libcairo2-dev \
-		libcutensor1 \
-		libcutensor-dev \
+		libcutensor1=1.6.1.5-1 \
+		libcutensor-dev=1.6.1.5-1 \
 		libglib2.0-dev \
 		libgstreamer1.0-0 \
 		libgtk2.0-dev \
@@ -55,6 +57,7 @@ RUN set -exuo pipefail ; \
 		time \
 		unzip \
 		wget \
+		cuda-drivers=520.61.05-1\
 	; \
 	pip3 install cython ; \
 	curl -L https://github.com/Kitware/ninja/releases/download/v1.10.0.gfb670.kitware.jobserver-1/ninja-1.10.0.gfb670.kitware.jobserver-1_x86_64-linux-gnu.tar.gz | tar xzv --strip-components=1 -C /usr/local/bin ; \
@@ -71,4 +74,4 @@ RUN set -exuo pipefail ; \
 	rm -r *.deb ccache*
 
 ENV CUDA_PATH=/usr/local/cuda
-ENV PATH=${CUDA_PATH}/bin:$PATH
+ENV PATH=${CUDA_PATH}/bin:$PATH
diff --git a/modules/nvidia_plugin/README.md b/modules/nvidia_plugin/README.md
@@ -7,9 +7,9 @@ The plugin uses custom kernels and [cuBLAS, cuDNN, cuTENSOR libraries\*] as a ba
 ## Supported Platforms
 OpenVINO™ NVIDIA GPU plugin is supported and validated on the following platforms: 
 
-OS                     | GPU
----------------------- | ----------------------
-Ubuntu* 20.04 (64-bit) | NVIDIA Quadro RTX 4000
+OS                     | GPU                   | Driver               |
+---------------------- | --------------------- |--------------------- |
+Ubuntu* 20.04 (64-bit) | NVIDIA Quadro RTX 4000| 520.61.05            |
 
 ## Distribution
 OpenVINO™ NVIDIA GPU plugin is not included into Intel® Distribution of OpenVINO™. To use the plugin, it should be built from source code.
@@ -32,13 +32,13 @@ sudo apt-get update
 sudo apt-get install clang-8 clang++8
 ```
 
-2. Install **NVIDIA 460** version of driver from [NVIDIA download drivers](http://www.nvidia.com/Download/index.aspx?lang=en-us)
-3. Install **CUDA 11.2** from [How to install CUDA](https://docs.nvidia.com/cuda/cuda-quick-start-guide/index.html)
+2. Install suitable **NVIDIA driver** from [NVIDIA download drivers](http://www.nvidia.com/Download/index.aspx?lang=en-us)
+3. Install **CUDA 11.8** from [How to install CUDA](https://docs.nvidia.com/cuda/cuda-quick-start-guide/index.html)
 
    Do not forget to add `<path_to_cuda>/bin/` in **PATH** variable for example `export PATH="<path_to_cuda>/bin:$PATH"`    
 
-4. Install **cuDNN 8.1.0** from [How to install cuDNN](https://docs.nvidia.com/deeplearning/cudnn/install-guide/index.html)
-5. Install **cuTENSOR 1.3.0** from [How to install cuTENSOR](https://docs.nvidia.com/cuda/cutensor/getting_started.html#installation-and-compilation)
+4. Install **cuDNN 8.6.0** from [How to install cuDNN](https://docs.nvidia.com/deeplearning/cudnn/install-guide/index.html)
+5. Install **cuTENSOR 1.6.1** from [How to install cuTENSOR](https://docs.nvidia.com/cuda/cutensor/getting_started.html#installation-and-compilation)
 
 ### Build with cmake
 
@@ -107,12 +107,12 @@ git checkout develop
 ```
 3. Setup `CUDACXX` environment variable to point to the CUDA _nvcc_ compiler like the next (use yours path)
 ```bash
-export CUDACXX=/usr/local/cuda-11.2/bin/nvcc
+export CUDACXX=/usr/local/cuda-11.8/bin/nvcc
 ```
 
 4. Add the path to the cuda libraries to the `LD_LIBRARY_PATH` environment variable like the next (use yours path)
 ```bash
-export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-11.2/bin/nvcc
+export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-11.8/bin/nvcc
 ```
 
 5. Run setup.py build command as follows.