nvidia graphic card 설치 및 Cuda & Cudnn 설치

nvidia graphic card 설치 

 

$ sudo apt-add-repository ppa:graphics-drivers/ppa -y
$ sudo apt-get update
$ sudo apt-cache search ^nvidia-driver
# $ ubuntu-drivers devices # 추천하는 드라이버 확인
$ sudo apt-get install nvidia-driver-460

(reboot)

$ nvidia-smi

 

 

 

cuda 설치 (toolkit 11.2)

 

graphic card 460을 다운로드 하였기 때문에 호환되는 버전인 
11.2를 다운로드 하게 되었다.

 

 

(버전 확인 사이트 )
https://docs.nvidia.com/deploy/cuda-compatibility/index.html
(다운로드 사이트 )
https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arc
h=x86_64&target_distro=Ubuntu&target_version=1804&target_type=runfi
lelocal
선택사항 : linux, x86_64, Ubuntu, 18.04, runfile(local)

 

(터미널에서 명령어 실행)

$wget 
https://developer.download.nvidia.com/compute/cuda/11.2.2/local_installer
s/cuda_11.2.2_460.32.03_linux.run
$sudo sh cuda_11.2.2_460.32.03_linux.run

(여기 아주 조심해야함 !!)

cuda toolkit 만을 제외하고 space bar를 통해 체크박스로 체크 
표시를 지우고, 설치 

설치후 $ sudo nano ~/.bashrc 라는 명령어 입력

 

export PATH=/usr/local/cuda-11.2/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-11.2/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

을 추가하고 Ctrl+O, Enter, Ctrl+x 순으로 입력한다. 

 

$ source ~/.bashrc 라는 명령어 입력 후 
$ nvcc -V 입력하고
버전이 맞는지 확인한다

 

cudnn 설치

https://developer.nvidia.com/rdp/cudnn-archive
(해당 사이트에서 내 cuda version인 11.2에 맞는 cuDNN 8.1.0 & Linux 
x86_64로 다운로드)
(download folder에 다운되어 있다.)

$ tar -xvzf cudnn-11.2-linux-x64-v8.1.0.77.tgz

$ sudo cp cuda/include/cudnn.h /usr/local/cuda-11.2/include
$ sudo cp cuda/include/cudnn_version.h /usr/local/cuda-11.2/include

$ sudo cp -P cuda/lib64/libcudnn* /usr/local/cuda-11.2/lib64
$ sudo chmod a+r /usr/local/cuda-11.2/include/cudnn.h 
/usr/local/cuda-11.2/lib64/libcudnn*

<confirmation>
$ cat /usr/local/cuda/include/cudnn_version.h | grep CUDNN_MAJOR 
-A 2