Ubuntu 18.04安装 CUDA 10.1 、cuDNN 7.6.5、PyTorch1.3

陈多喜 發表於 2019-11-11 19:45:00

Ubuntu 18.04安装 CUDA 10.1 、cuDNN 7.6.5、PyTorch1.3

转载请注明出处  BooTurbo https://www.cnblogs.com/booturbo/p/11834661.html
 
安装平台及环境
CPU：i9-9900k桌面级
GPU：RTX 2080移动版
系统：Ubuntu 18.04.3 LTS
1、在安装CUDA之前确保环境满足安装条件
2、进入NVIDIA官网下载适合自己机器的CUDA版本，官网下载，如图所示，按照 Installation Instructions 来进行，
<div class="cnblogs_code">
<pre>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 http://developer.download.nvidia.com/compute/cuda/10.1/Prod/local_installers/cuda-repo-ubuntu1804-10-1-local-10.1.243-418.87.00_1.0-1_amd64.deb

sudo dpkg -i cuda-repo-ubuntu1804-10-1-local-10.1.243-418.87.00_1.0-1_amd64.deb

sudo apt-key add /var/cuda-repo-10-1-local-10.1.243-418.87.00/7fa2af80.pub

sudo apt-get update sudo apt-get -y install cuda</pre>
</div>
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191111101307549-1660383844.png">
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191111102317917-1464934472.png">
3、安装完成后，添加环境变量，打开 bashrc 文件，
命令行输入： sudo gedit ~/.bashrc ，然后在文件最后添加下面3行，保存
<div class="cnblogs_code">
<pre>export CUDA_HOME=/usr/local/cuda

export PATH=$PATH:$CUDA_HOME/bin

export LD_LIBRARY_PATH=/usr/local/cuda-10.1/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}} </pre>
</div>
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191111104850204-72975851.png">
然后刷新下环境变量， 
输入： source ~/.bashrc 
4、测试下CUDA是否安装成功，
方法1：输入 nvcc -V ，显示如下，说明没问题
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191111104633098-2058118465.png">
方法2：输入如下，显示如图说明安装成功
<div class="cnblogs_code">
<pre>cd /usr/local/cuda/samples/1_Utilities/deviceQuery
sudo make
./deviceQuery</pre>
</div>
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191111110057169-2068921376.png">
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191111110109300-551046699.png">
5、安装cuDNN 7.6.5版本，转到官网下载，下载前先注册一下，填个调查问卷，根据自己的环境和架构选择包，下载到本地
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191111123807202-1429904716.png">
切换到下载目录 cd Downloads ，然后按照以下操作进行，
解压 cuDNN Library for Linux，输入：  tar -zxvf cudnn-10.1-linux-x64-v7.6.5.32.tgz 
将解压出来的文件复制到安装的CUDA环境中，输入： sudo cp cuda/include/cudnn.h /usr/local/cuda/inlude 
                                                                                                             sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64   
更改权限，输入： sudo chmod a+r /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn*  
安装 Deb 包，cuDNN Runtime Library for Ubuntu18.04(Deb)，cuDNN Developer Library for Ubuntu18.04(Deb)，cuDNN Code Samples and User Guide for Ubuntu18.04(Deb)
分别输入：   sudo dpkg -i libcudnn7_7.6.5.32-1+cuda10.1_amd64.deb 
                       sudo dpkg -i libcudnn7-dev_7.6.5.32-1+cuda10.1_amd64.deb  
                       sudo dpkg -i libcudnn7-doc_7.6.5.32-1+cuda10.1_amd64.deb  
安装结束后，重启系统，再测试一下安装是否成功，
方法1，输入： cp -r /usr/src/cudnn_samples_v7/ ~ 
                           cd ~/cudnn_samples_v7/mnistCUDNN 
                           make clean && make 
                           ./mnistCUDNN 
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191111140553030-1221845967.png">
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191111140642991-783440879.png">
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191111140811958-505170999.png">
出现Test passed！没有报错即安装成功
方法2，输入： cd /usr/local/cuda/samples/1_Utilities/deviceQuery 
                           sudo make 
                           ./deviceQuery 
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191111141949508-789496593.png">
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191111142002339-1894458334.png">
出现PASS结果，说明安装成功。
 6、安装PyTorch1.3
进入PyTorch官网查找合适的版本，官网 ，
输入： pip3 install torch torchvision ，等待安装结束
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191112092700262-318991109.png">
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191112092023413-1340168004.png">
 安装完成后，进入python3环境，验证是否成功，
输入： python3，
再输入： import torch
                import torchvision
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191112093611581-1188649008.png">
 没有报错，说明安装成功。
最后验证下GPU能否使用，输入： print(torch.cuda.is_available()) ，输出True，说明没问题。
<img src="https://img2018.cnblogs.com/common/1585117/201911/1585117-20191112094442625-737983084.png">
 
 
Enjoy it.
  
来源：https://www.cnblogs.com/booturbo/p/11834661.html

頁: [1]

琼殿技术论坛's Archiver

Ubuntu 18.04安装 CUDA 10.1 、cuDNN 7.6.5、PyTorch1.3