ubuntu20.04安装/重装 cuda11.4、cudnn

慎重安装最新版的cuda吧，看看当前版本的pytorch和tensorflow支不支持最新的cuda，最好选个两个都支持的cuda版本，安装流程是一样的

1.检查自己电脑支持的cuda

lhw@lhw-Dell-G15-5511:~$ nvidia-smi
Wed Oct 20 00:00:21 2021       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.63.01    Driver Version: 470.63.01    CUDA Version: 11.4     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|                               |                      |               MIG M. |
|===============================+======================+======================|
|   0  NVIDIA GeForce ...  Off  | 00000000:01:00.0  On |                  N/A |
| N/A   49C    P8    16W /  N/A |    633MiB /  5938MiB |     16%      Default |
|                               |                      |                  N/A |
+-------------------------------+----------------------+----------------------+
                                                                               
+-----------------------------------------------------------------------------+
| Processes:                                                                  |
|  GPU   GI   CI        PID   Type   Process name                  GPU Memory |
|        ID   ID                                                   Usage      |
|=============================================================================|
|    0   N/A  N/A       988      G   /usr/lib/xorg/Xorg                 35MiB |
|    0   N/A  N/A      1586      G   /usr/lib/xorg/Xorg                345MiB |
|    0   N/A  N/A      1724      G   /usr/bin/gnome-shell               57MiB |
|    0   N/A  N/A    164717      G   ...AAAAAAAAA= --shared-files      144MiB |
|    0   N/A  N/A    164726      G   ...AAAAAAAAA= --shared-files       38MiB |
+-----------------------------------------------------------------------------+

显示支持的cuda版本为11.4

查看显卡算力地址 

 

2.去nvidia官网下载最新的cuda11.4.2

 

wget https://developer.download.nvidia.com/compute/cuda/11.4.2/local_installers/cuda_11.4.2_470.57.02_linux.run
sudo sh cuda_11.4.2_470.57.02_linux.run

出来提示后，依次操作（如果已经安装了NVIDIA驱动就把第一项的x驱动安装去掉）

 1.是否接受EULA？   输入accept
 2.cuda安装？   选择不安装驱动，其他默认安装install
 直至安装完成

添加环境变量：

gedit ~/.bashrc
# 文本最后添加以下内容：
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64
export PATH=$PATH:/usr/local/cuda/bin
export CUDA_HOME=$CUDA_HOME:/usr/local/cuda
# 保存退出，打开新终端激活
source ~/.bashrc

测试CUda

cd /usr/local/cuda/samples/1_Utilities/deviceQuery
sudo make -j4
./deviceQuery

显示结果：

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 11.4, CUDA Runtime Version = 11.4, NumDevs = 1
Result = PASS

 

3.安装cuDNN

去官网注册账号登录下载cuDNN

cuDNN Library for Linux (x86_64)

下载后解压复制

tar zxvf cudnn-11.4-linux-x64-v8.2.4.15.tgz

sudo cp cuda/include/cudnn* /usr/local/cuda/include
sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64
sudo chmod a+r /usr/local/cuda/include/cudnn* /usr/local/cuda/lib64/libcudnn*

  

 

查询版本号

cat cuda/include/cudnn_version.h |grep ^#

#ifndef CUDNN_VERSION_H_
#define CUDNN_VERSION_H_
#define CUDNN_MAJOR 8
#define CUDNN_MINOR 2
#define CUDNN_PATCHLEVEL 4
#define CUDNN_VERSION (CUDNN_MAJOR * 1000 + CUDNN_MINOR * 100 + CUDNN_PATCHLEVEL)
#endif /* CUDNN_VERSION_H */

 

4. cuda 卸载

 sudo apt autoremove cuda

cd /usr/local/cuda/bin/
sudo ./cuda-uninstaller
# 选中所有cuda相关选项
sudo rm -rf /usr/local/cuda-11.0
sudo rm -rf /usr/local/cuda

5.  cudnn卸载

sudo rm -rf /usr/local/cuda/include/cudnn.h
sudo rm -rf /usr/local/cuda/lib64/libcudnn*

 

 

 

参考博客1

参考博客2