在官网选择需要的版本

选择安装途径，选择runfile

需要降低C++的版本，否则

执行sudo sh cuda_11.1.0_455.23.05_linux.run,会出现

查看对应的文件，会有

可以加上override参数之后，继续安装

降低gcc版本，但是会出现错误

E: Package ‘g+±7’ has no installation candidate

解决E: Package ‘g+±7’ has no installation candidate错误，需要更换ubuntun 20.04的镜像源

# 修改 apt-get 源的配置文件
sudo cp /etc/apt/sources.list /etc/apt/sources.list.bak #备份
sudo vim /etc/apt/sources.list #修改

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


#最后执行
sudo apt-get update

如果删除了原本的镜像源内容，使用nvcc -V命令，发现没有了，安装的话也会失败

The following packages have unmet dependencies:  nvidia-cuda-toolkit :
Depends: nvidia-cuda-dev (= 10.1.243-3) but it is not going to be
installed E: Unable to correct problems, you have held broken
packages.
Running `sudo apt install nvidia-cuda-dev` yields:

The following packages have unmet dependencies:  nvidia-cuda-dev :
Depends: libcublas10 (= 10.1.243-3) but 10.2.2.214-1 is to be
installed
                   Recommends: libvdpau-dev but it is not going to be installed
                   Recommends: libnvcuvid1 but it is not installable E: Unable to correct problems, you have held broken packages.

原因更改了 /etc/apt/sources.list can’t install nvidia-driver-toolkit on Ubuntu 20.04 LTS - needs uninstallable package - Ask Ubuntu

方法：在/etc/apt/sources.list中添加

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

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

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

镜像源更改之后，执行

sudo apt-cache policy
sudo apt-get --purge remove "*cublas*" "cuda*" "*nvidia*"
sudo apt-get clean
sudo apt-get autoremove
sudo apt-get update
sudo apt-get upgrade
sudo apt-get install nvidia-cuda-toolkit

再次执行安装nvcc的命令

降低g++版本之后，可以正常执行 sudo sh cuda_11.1.0_455.23.05_linux.run, 前面有X的是选中要安装的，[ ]是不安装的

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-KgFx3ljk-1681261652591)(E:\Linux\WSL\wsl安装torch_geometric.assets\image-20230411171049593.png)]

安装完成后显示

正常安装cuda 11.1了，虽然nvidia-smi显示和nvcc -V显示的版本不同，以nvcc -V为主

/usr/local目录下有了 cuda目录

更新可执行文件路径,在 vim ~/.bashrc中加入,如果不在 ~/.bashrc中加入的话，下次启动还是找不到cuda

export PATH=/usr/local/cuda-11.1/bin${PATH:+:${PATH}}

更新动态链接库路径，在 vim ~/.bashrc中加入

export LD_LIBRARY_PATH=/usr/local/cuda-11.1/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

安装1.8.1版本的torch https://pytorch.org/get-started/previous-versions/

# CUDA 11.1
pip install torch==1.8.1+cu111 torchvision==0.9.1+cu111 torchaudio==0.8.1 -f https://download.pytorch.org/whl/torch_stable.html

然后安装torch-geometric, 根据torch和cuda版本选择，

安装顺序为 torch-scatter2.0.8，torch-sparse0.6.12，torch-cluster1.5.9，torch-spline-conv1.2.1，torch-geometric==1.7.0，我的python版本为3.6， wsl(linux)