linux Ubuntu Server 20.04 系统安装（四）：深度学习 GPU 环境配置（CUDA12.2驱动+CUDA Toolkit 12.2+cuDNN v8.9.7）

引言

本文以 Ubuntu 20.04 操作系统为例，演示如何配置深度学习 GPU 环境。

一、NVIDIA 显卡驱动的安装

访问如下网址

https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&Distribution=Ubuntu&target_version=20.04&target_type=runfile_local

下载推荐的cuda并安装

复制箭头所指处的命令到命令行

wget https://developer.download.nvidia.com/compute/cuda/12.2.0/local_installers/cuda_12.2.0_535.54.03_linux.run

必须先关图形界面（重要）

如果不关，运行下面安装的命令之后会报错，装不上

systemctl isolate multi-user.target

关闭图形界面之后会重启，输入如下命令

安装驱动

sudo sh cuda_12.2.0_535.54.03_linux.run

稍等一会在弹出的安装界面中选“continue”，如果选了会跳出安装，就说明安装失败，给了失败日志的路径，自己查看原因，一般是gcc版本问题，

gcc降级

sudo apt-get install gcc-7 g++-7

sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-7 9

sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-9 1

sudo update-alternatives --display gcc //查看默认版本

sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-7 9

sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-9 1

sudo update-alternatives --display g++

如果没问题继续安装： cuda安装包是自带显卡驱动的，所以这一步按空格去掉安装显卡驱动的选项，然后选择install

开启图形界面

sudo systemctl start graphical.target

二、安装 CUDA Toolkit 12.2

准备

开始安装之前，需要修改 apt-get 的镜像源，不然国内下载速度很慢。我们这里以阿里源为例演示如何修改。

备份官方源

sudo mv /etc/apt/sources.list /etc/apt/sources.list.bak

安装vim：

sudo apt-get install vim

运行 sudo vim /etc/apt/sources.list 打开文件，并按 i 插入以下内容，按 ESC 后再按 :wq 保存退出

sudo vim /etc/apt/sources.list

deb http://mirrors.aliyun.com/ubuntu/ focal main restricted universe multiverse

deb-src http://mirrors.aliyun.com/ubuntu/ focal main restricted universe multiverse

deb http://mirrors.aliyun.com/ubuntu/ focal-security main restricted universe multiverse

deb-src http://mirrors.aliyun.com/ubuntu/ focal-security main restricted universe multiverse

deb http://mirrors.aliyun.com/ubuntu/ focal-updates main restricted universe multiverse

deb-src http://mirrors.aliyun.com/ubuntu/ focal-updates main restricted universe multiverse

# deb http://mirrors.aliyun.com/ubuntu/ focal-proposed main restricted universe multiverse

# deb-src http://mirrors.aliyun.com/ubuntu/ focal-proposed main restricted universe multiverse

deb http://mirrors.aliyun.com/ubuntu/ focal-backports main restricted universe multiverse

deb-src http://mirrors.aliyun.com/ubuntu/ focal-backports main restricted universe multiverse

更新数据源列表

sudo apt-get update

安装参考 NVIDIA 官网提供的教程。

Quick Start Guide 12.3 documentation

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-ubuntu2004.pin

sudo mv cuda-ubuntu2004.pin /etc/apt/preferences.d/cuda-repository-pin-600

wget https://developer.download.nvidia.com/compute/cuda/12.3.1/local_installers/cuda-repo-ubuntu2004-12-3-local_12.3.1-545.23.08-1_amd64.deb

sudo dpkg -i cuda-repo-ubuntu2004-12-3-local_12.3.1-545.23.08-1_amd64.deb

sudo cp /var/cuda-repo-ubuntu2004-12-3-local/cuda-*-keyring.gpg /usr/share/keyrings/

sudo apt-get update

sudo apt-get -y install cuda-toolkit-12-3

sudo reboot # 非必需

执行完上述安装操作后，需要配置环境变量，让该系统下的所有用户都能使用：

打开全局配置文件

sudo vi /etc/profile

在文件末尾添加如下内容后保存退出。

export PATH=/usr/local/cuda-12.3/bin${PATH:+:${PATH}}

export LD_LIBRARY_PATH=/usr/local/cuda-12.3/lib64\${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

激活环境变量

source /etc/profile

运行 nvcc -V 来校验安装是否成功，若正确显示版本号则说明安装成功

$ nvcc -V

nvcc: NVIDIA (R) Cuda compiler driver

Built on Fri_Nov__3_17:16:49_PDT_2023

Cuda compilation tools, release 12.3, V12.3.103

Build cuda_12.3.r12.3/compiler.33492891_0

二、安装 cuDNN v8.9.7

安装参考 NVIDIA 官方提供的教程。

从官网下载合适版本的离线安装包

2. 开始安装

sudo dpkg -i cudnn-local-repo-ubuntu2004-8.9.7.29_1.0-1_amd64.deb

sudo cp /var/cudnn-local-repo-ubuntu2004-8.9.7.29/cudnn-local-30472A84-keyring.gpg /usr/share/keyrings/

cd /var/cudnn-local-repo-ubuntu2004-8.9.7.29/

sudo dpkg -i libcudnn8_8.9.7.29-1+cuda12.2_amd64.deb

sudo dpkg -i libcudnn8-dev_8.9.7.29-1+cuda12.2_amd64.deb

sudo dpkg -i libcudnn8-samples_8.9.7.29-1+cuda12.2_amd64.deb

校验安装是否成功（最终输出 Test passed! 则说明安装成功）

cp -r /usr/src/cudnn_samples_v8/ $HOME

cd $HOME/cudnn_samples_v8/mnistCUDNN

make clean && make

./mnistCUDNN

校验安装时如果遇到下述错误：

```bash

test.c:1:10: fatal error: FreeImage.h: No such file or directory

1 | #include "FreeImage.h"

| ^~~~~~~~~~~~~

compilation terminated.

>>> WARNING - FreeImage is not set up correctly. Please ensure FreeImage is set up correctly. <<<

可通过如下方式解决，然后重新进行校验

sudo apt-get -y install libfreeimage3 libfreeimage-dev

重启后验证通过

sudo reboot

验证通过：

wyr@ubuntu2004w1:~/cudnn_samples_v8/mnistCUDNN$ ./mnistCUDNN

Executing: mnistCUDNN

cudnnGetVersion() : 8401 , CUDNN_VERSION from cudnn.h : 8401 (8.4.1)

Host compiler version : GCC 9.4.0

There are 2 CUDA capable devices on your machine :

device 0 : sms 56 Capabilities 6.0, SmClock 1328.5 Mhz, MemSize (Mb) 16280, MemClock 715.0 Mhz, Ecc=1, boardGroupID=0

device 1 : sms 56 Capabilities 6.0, SmClock 1328.5 Mhz, MemSize (Mb) 16280, MemClock 715.0 Mhz, Ecc=1, boardGroupID=1

Using device 0

Testing single precision

Loading binary file data/conv1.bin

Loading binary file data/conv1.bias.bin

Loading binary file data/conv2.bin

Loading binary file data/conv2.bias.bin

Loading binary file data/ip1.bin

Loading binary file data/ip1.bias.bin

Loading binary file data/ip2.bin

Loading binary file data/ip2.bias.bin

Loading image data/one_28x28.pgm

Performing forward propagation ...

Testing cudnnGetConvolutionForwardAlgorithm_v7 ...

^^^^ CUDNN_STATUS_SUCCESS for Algo 1: -1.000000 time requiring 0 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 0: -1.000000 time requiring 0 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 2: -1.000000 time requiring 57600 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 5: -1.000000 time requiring 178432 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 4: -1.000000 time requiring 184784 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 7: -1.000000 time requiring 2057744 memory

^^^^ CUDNN_STATUS_NOT_SUPPORTED for Algo 6: -1.000000 time requiring 0 memory

^^^^ CUDNN_STATUS_NOT_SUPPORTED for Algo 3: -1.000000 time requiring 0 memory

Testing cudnnFindConvolutionForwardAlgorithm ...

^^^^ CUDNN_STATUS_SUCCESS for Algo 1: 0.024864 time requiring 0 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 0: 0.025152 time requiring 0 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 2: 0.066912 time requiring 57600 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 5: 0.097024 time requiring 178432 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 7: 0.144352 time requiring 2057744 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 4: 0.275520 time requiring 184784 memory

^^^^ CUDNN_STATUS_NOT_SUPPORTED for Algo 6: -1.000000 time requiring 0 memory

^^^^ CUDNN_STATUS_NOT_SUPPORTED for Algo 3: -1.000000 time requiring 0 memory

Testing cudnnGetConvolutionForwardAlgorithm_v7 ...

^^^^ CUDNN_STATUS_SUCCESS for Algo 1: -1.000000 time requiring 0 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 0: -1.000000 time requiring 0 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 2: -1.000000 time requiring 128000 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 5: -1.000000 time requiring 4656640 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 7: -1.000000 time requiring 1433120 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 4: -1.000000 time requiring 2450080 memory

^^^^ CUDNN_STATUS_NOT_SUPPORTED for Algo 6: -1.000000 time requiring 0 memory

^^^^ CUDNN_STATUS_NOT_SUPPORTED for Algo 3: -1.000000 time requiring 0 memory

Testing cudnnFindConvolutionForwardAlgorithm ...

^^^^ CUDNN_STATUS_SUCCESS for Algo 1: 0.062560 time requiring 0 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 0: 0.062880 time requiring 0 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 4: 0.107712 time requiring 2450080 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 7: 0.114528 time requiring 1433120 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 2: 0.116704 time requiring 128000 memory

^^^^ CUDNN_STATUS_SUCCESS for Algo 5: 0.130944 time requiring 4656640 memory

^^^^ CUDNN_STATUS_NOT_SUPPORTED for Algo 6: -1.000000 time requiring 0 memory

^^^^ CUDNN_STATUS_NOT_SUPPORTED for Algo 3: -1.000000 time requiring 0 memory

Resulting weights from Softmax:

0.0000000 0.9999399 0.0000000 0.0000000 0.0000561 0.0000000 0.0000012 0.0000017 0.0000010 0.0000000

Loading image data/three_28x28.pgm