概念介绍

NVIDIA 驱动：显卡驱动，安装后可以使用 nvidia-smi 命令。
CUDA：CUDA是一个并行计算平台和编程模型，能够使得使用GPU进行通用计算变得简单和优雅。可以看作编程语言，也可以看作 API。
- 运行CUDA应用程序要求系统至少具有一个具有CUDA功能的GPU和与CUDA Toolkit兼容的驱动程序。
CUDA Toolkit (NVIDIA)： CUDA完整的工具安装包，其中提供了 Nvidia 驱动程序、开发 CUDA 程序相关的开发工具包等可供安装的选项。包括 CUDA 程序的编译器、IDE、调试器等，CUDA 程序所对应的各式库文件以及它们的头文件。
- CUDA Toolkit 有的会包含显卡驱动。
CUDA Toolkit (Pytorch)： CUDA不完整的工具安装包，其主要包含在使用 CUDA 相关的功能时所依赖的动态链接库。不会安装驱动程序。只能运行编译好的 CUDA 程序。
NVCC：CUDA的编译器，只是 CUDA Toolkit 中的一部分。
cuDNN：为深度学习计算设计的软件库。是NVIDIA专门针对深度神经网络中的基础操作而设计基于GPU的加速库。cuDNN为深度神经网络中的标准流程提供了高度优化的实现方式，例如convolution、pooling、normalization以及activation layers的前向以及后向过程。

CUDA 环境安装

服务器基础环境安装

基础软件

sudo apt-get update
sudo apt-get upgrade -y
sudo apt-get install net-tools -y
sudo apt-get install tree -y
sudo apt-get install vim -y
sudo apt-get install gcc -y
sudo apt-get install g++ -y

SSH 和 SFTP

sudo apt-get install openssh-server -y

sudo apt-get install vsftpd
sudo vim /etc/vsftpd.conf

'''
local_enable=YES
write_enable=YES
'''

sudo /etc/init.d/vsftpd restart

Anaconda

# https://repo.anaconda.com/archive/
wget https://repo.anaconda.com/archive/Anaconda3-5.3.1-Linux-x86_64.sh 

chmod +x Anaconda3-5.3.1-Linux-x86_64.sh 
bash Anaconda3-5.3.1-Linux-x86_64.sh 

sudo vim ~/.bashrc

'''
export PATH="/home/用户名/anaconda3/bin:$PATH"
'''

source ~/.bashrc

conda config --set ssl_verify false

替换 apt 源

sudo cp /etc/apt/sources.list /etc/apt/sources.list.bak 
sudo rm -rf /etc/apt/sources.list
sudo vim /etc/apt/sources.list

'''
deb http://mirrors.aliyun.com/ubuntu/ trusty main restricted universe multiverse
deb http://mirrors.aliyun.com/ubuntu/ trusty-security main restricted universe multiverse
deb http://mirrors.aliyun.com/ubuntu/ trusty-updates main restricted universe multiverse
deb http://mirrors.aliyun.com/ubuntu/ trusty-proposed main restricted universe multiverse
deb http://mirrors.aliyun.com/ubuntu/ trusty-backports main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ trusty main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ trusty-security main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ trusty-updates main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ trusty-proposed main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ trusty-backports main restricted universe multiverse
'''

sudo apt-get update

Git

sudo apt-get install git -y
ssh-keygen
cat /home/用户名/.ssh/id_rsa.pub

NVIDIA 驱动安装

添加 nvidia repository

sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt update

选择驱动版本并安装

sudo ubuntu-drivers devices

显示可用的驱动版本，例如：

driver   : nvidia-410 - third-party free
driver   : nvidia-415 - third-party free
driver   : nvidia-418 - third-party free
driver   : nvidia-384 - distro non-free
driver   : nvidia-430 - third-party free recommended
driver   : xserver-xorg-video-nouveau - distro free builtin

安装选定的版本，410版本为例，或者自动安装推荐版本

sudo ubuntu-drivers autoinstall
sudo apt install nvidia-410

重启，然后通过用如下命令查看显卡信息

nvidia-smi

watch -n 1 nvidia-smi

nvidia-smi -l 1

不建议使用watch查看nvidia-smi，watch每个时间周期开启一个进程(PID)，查看后关闭进程，会影响cuda操作，如cudaMalloc；建议使用nvidia-smi -l x或者nvidia-smi --loop=xxx代替，这个命令执行期间一直是一个进程PID。

CUDA-toolkit 安装

# https://developer.nvidia.com/zh-cn/cuda-toolkit

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-ubuntu1804.pin
sudo mv cuda-ubuntu1804.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/11.6.2/local_installers/cuda-repo-ubuntu1804-11-6-local_11.6.2-510.47.03-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu1804-11-6-local_11.6.2-510.47.03-1_amd64.deb
sudo apt-key add /var/cuda-repo-ubuntu1804-11-6-local/7fa2af80.pub
sudo apt-get update
sudo apt-get -y install cuda


wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-ubuntu1804.pin
sudo mv cuda-ubuntu1804.pin /etc/apt/preferences.d/cuda-repository-pin-600
sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub
sudo add-apt-repository "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/ /"
sudo apt-get update
sudo apt-get -y install cuda

reboot
nvidia-smi

vim ~/.bashrc

'''
export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
'''

source ~/.bashrc
nvcc -V

https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&Distribution=Ubuntu&target_version=18.04&target_type=deb_network

cuDNN 安装

# https://developer.nvidia.com/rdp/cudnn-download

wget https://developer.nvidia.com/compute/cudnn/secure/8.4.0/local_installers/10.2/cudnn-local-repo-ubuntu1804-8.4.0.27_1.0-1_amd64.deb
sudo dpkg -i cudnn-local-repo-ubuntu1804-8.4.0.27_1.0-1_amd64.deb