ubuntu에서 cuda, cudnn, tensorflow 셋업하기

Target

UBUNTU 22.04
CUDA = 11.08
CUDNN = 8.6.0
Tensorflow = 2.12.0

tensorflow install

pip3 install tensorflow==2.12.0

CUDA install

linstall

wget https://developer.download.nvidia.com/compute/cuda/11.8.0/local_installers/cuda_11.8.0_520.61.05_linux.run
sudo sh cuda_11.8.0_520.61.05_linux.run

/usr/local/cuda-11.8 에 압축이 풀림
/usr/local/cuda에 심볼릭 링킹됨. 

$vim ~/.bashrc 아래 추가
export PATH="/usr/local/cuda-11.8/bin:$PATH"
export LD_LIBRARY_PATH="/usr/local/cuda-11.8/lib64:$LD_LIBRARY_PATH"

$source ~/.bashrc

nvcc -V로 버젼 확인 가능 함.

CUDNN install

download link
-> Download cuDNN v8.6.0 (October 3rd, 2022), for CUDA 11.x -» Local Installer for Linux x86_64 (Tar)

install guide

2.3.1. Tar File Installation
Before issuing the following commands, you'll need to replace x.x and v8.x.x.x with your specific CUDA and cuDNN versions and package date.
Procedure
Navigate to your <cudnnpath> directory containing the cuDNN tar file.
Unzip the cuDNN package.
$ tar -xvf cudnn-linux-x86_64-8.6.0.163_cuda11-archive.tar.xz cuda

Copy the following files into the CUDA Toolkit directory.
$ sudo cp cuda/include/cudnn*.h /usr/local/cuda/include 
$ sudo cp -P cuda/lib/libcudnn* /usr/local/cuda/lib64 
$ sudo chmod a+r /usr/local/cuda/include/cudnn*.h /usr/local/cuda/lib64/libcudnn*

cudnn version 확인

cuDNN 8.x 이전은  cat /usr/local/cuda/include/cudnn.h 파일에서 확인을 해야 합니다.
cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2
cuDNN 8.x 이후 버전에서는 위의 명령어로 설치 확인이 되지 않습니다.

cuDNN 8.x 이상 부터는 cudnn_version.h 파일에서 버전을 확인해야 합니다.
cat /usr/local/cuda/include/cudnn_version.h | grep CUDNN_MAJOR -A 2