cuda卸载与安装-Toy模板网

这篇具有很好参考价值的文章主要介绍了cuda卸载与安装。希望对大家有所帮助。如果存在错误或未考虑完全的地方，请大家不吝赐教，您也可以点击"举报违法"按钮提交疑问。

前言

用deb方式安装的cuda，进行卸载。我目前是卸载10.1，安装10.2或者11.6。
CUDA与显卡型号是对应的，若显卡升级了，那CUDA也要跟着升级，不然无法适配显卡，找不到设备。

RTX2060s/RTX2070s 可以对用CUDA11.6，cudnn8.4.0, Driver Version 510
RTX3050 用CUDA11.6就不行了。先安装software&update中适配的驱动，然后在完善相关的CUDA和CUDNN

显卡基本信息

cuda卸载与安装

一、卸载

参考的方法完全卸载cuda

参考problem-while-installing-cuda-toolkit-in-ubuntu-18-04

1.1 首先关闭使用NVIDIA的驱动程序

You have to uninstall any nvidia driver before running sudo apt install -y cuda
To do so, got to "Software & Updates" -> "Additional drivers" -> Using X.Org X (nouveou)

1.2 清除cuda，另外

sudo apt clean
sudo apt update
sudo apt purge nvidia-* 
sudo apt autoremove

旧方法(2023.11.15推荐)，先使用sudo apt-get --purge remove,再逐个文件夹删除，也清除得很彻底

第一步
sudo apt-get --purge remove cuda

# 自动删除一些没有依赖的cuda文件（此处注意不要删掉ros相关，以及内核文件相关,要么就不执行这一句，避免误删）
# 我尝试了下，果然把ros相关的也删除了，不过roscore还能用，关键是我希望把cuda安装好，ros之类可以重新装
sudo apt autoremove

# 删除cuda-10.1/11.6文件夹
sudo rm -rf /usr/local/cuda
sudo rm -rf /usr/local/cuda-*

# 查看关联文件，逐个删除
sudo find / -name cuda-*

# 尤其注意如下几个文件位置
sudo rm -rf /var/cuda*
sudo rm -rf /var/cudnn*
sudo rm -rf /usr/share/doc/cuda*
sudo rm -rf /usr/share/doc/cudnn*
sudo rm /etc/apt/sources.list.d/cuda*
sudo rm /etc/apt/sources.list.d/cudnn*
sudo rm /usr/lib/pkgconfig/cuda*

# 自动删除
sudo apt-get autoclean


# 注意
如果使用上面`sudo rm /etc/apt/sources.list.d/cuda*`会让`sudo apt update`时找不到cuda的源。需要逐个移除
sudo dpkg -l cuda # tab一下
sudo dpkg -P cuda # tab一下


# 删除cuDNN
sudo apt-get --purge remove libcudnn8
cd /usr/local/include/
sudo rm cudnn.h 

cd /usr/local/lib/
sudo rm libcudnn.*

# 删除TensorRT
sudo apt-get --purge remove nv-tensorrt-repo-ubuntu1804-cuda10.2-trt7.1.3.4-ga-20200617
sudo rm /etc/apt/sources.list.d/nv-tensorrt-cuda10.2-trt7.1.3.4-ga-20200617.list.save

# 配置文件 
cd  /etc/ld.so.conf.d/
# 看下哪些和cuda有关？都删掉 
sudo rm /etc/ld.so.conf.d/cuda-10-1.conf
# 旧的更新地址
sudo rm /etc/apt/sources.list.d/cuda-10-1-local-10.1.243-418.87.00.list.save
# 旧的dpkg信息
sudo rm /var/lib/dpkg/info/cuda-*

二、安装

先用这个选定显卡驱动，然后让它自动安装显卡驱动
cuda卸载与安装

接下来再进入官网，按照最新的安装命令，这里是cuda历史版本
https://developer.nvidia.com/cuda-toolkit-archive
下面是针对ubuntu18的，如果是20.04，则参考官网
2. 下载cuda-10.2

# 网址
[cuda-10.2 official link]](https://developer.nvidia.com/cuda-10.2-download-archive?target_os=Linux&target_arch=x86_64&target_distro=Ubuntu&target_version=1804&target_type=deblocal)
# 按照官网指令执行
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/10.2/Prod/local_installers/cuda-repo-ubuntu1804-10-2-local-10.2.89-440.33.01_1.0-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu1804-10-2-local-10.2.89-440.33.01_1.0-1_amd64.deb
sudo apt-key add /var/cuda-repo-10-2-local-10.2.89-440.33.01/7fa2af80.pub
sudo apt-get update
sudo apt-get -y install cuda
# 安装补丁
sudo dpkg -i cuda-repo-ubuntu1804-10-2-local_10.2.1-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu1804-10-2-local_10.2.2-1_amd64.deb

2.下载cuda-11-6

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/11.6.0/local_installers/cuda-repo-ubuntu1804-11-6-local_11.6.0-510.39.01-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu1804-11-6-local_11.6.0-510.39.01-1_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

下载cudnn-7.6.5 （对应cuda-10-2）

# 网址
https://developer.nvidia.com/rdp/cudnn-archive

# 找到这三个文件
# 包含了某些深度学习应用
# cuDNN Runtime Library for Ubuntu18.04 (Deb)
sudo dpkg -i libcudnn7_7.6.5.32-1+cuda10.2_amd64.deb
# 包含了在Ubuntu系统上开发深度学习时所需的cuDNN头文件
# cuDNN Developer Library for Ubuntu18.04 (Deb)
sudo dpkg -i libcudnn7-dev_7.6.5.32-1+cuda10.2_amd64.deb
# 包含了cuDNN例程和相关文档
# cuDNN Code Samples and User Guide for Ubuntu18.04 (Deb)
sudo dpkg -i libcudnn7-doc_7.6.5.32-1+cuda10.2_amd64.deb

3.下载cudnn-8.4.0（对应cuda-11-6）

# 网址
https://developer.nvidia.com/rdp/cudnn-archive
# 点击下载，只有一个文件
https://developer.nvidia.com/compute/cudnn/secure/8.4.0/local_installers/11.6/cudnn-local-repo-ubuntu1804-8.4.0.27_1.0-1_amd64.deb
# 安装
sudo dpkg -i cudnn-local-repo-ubuntu1804-8.4.0.27_1.0-1_amd64.deb
# 更新一下
sudo apt-get update
# 安装依赖
sudo apt-get install libcudnn8=8.4.0.27-1+cuda11.6
sudo apt-get install libcudnn8-dev=8.4.0.27-1+cuda11.6
sudo apt-get install libcudnn8-samples=8.4.0.27-1+cuda11.6

注意： cudnn-8.5.0对应cuda-11-7
3.1 验证安装文章来源地址https://www.toymoban.com/news/detail-415280.html

cp -r /usr/src/cudnn_samples_v8/ $HOME
cd  $HOME/cudnn_samples_v8/mnistCUDNN
make clean && make
./mnistCUDNN

配置环境变量

# 配置环境变量
 sudo gedit  ~/.zshrc
 
export PATH=/usr/local/cuda-10.2/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-10.2/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
 
# 更新环境变量
source ~/.zshrc
 
# 查看cuda是否安装成功
nvcc -V
# 我的显示如下
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Wed_Oct_23_19:24:38_PDT_2019
Cuda compilation tools, release 10.2, V10.2.89

重启电脑，查看是否正常

# 输入命令
$ nvidia-smi
# 显示
Wed Jan  4 11:06:49 2023       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.33.01    Driver Version: 440.33.01    CUDA Version: 10.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce RTX 207...  On   | 00000000:01:00.0  On |                  N/A |
| 40%   24C    P8    24W / 215W |    532MiB /  7979MiB |      1%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      1102      G   /usr/lib/xorg/Xorg                            56MiB |
|    0      1159      G   /usr/bin/gnome-shell                          49MiB |
|    0      2326      G   /usr/lib/xorg/Xorg                           259MiB |
|    0      2481      G   /usr/bin/gnome-shell                         156MiB |
|    0      2525      G   /opt/teamviewer/tv_bin/TeamViewer              2MiB |
|    0      2938      G   gnome-control-center                           2MiB |
+-----------------------------------------------------------------------------+