ubuntu安装NVIDIA驱动+CUDA+CUDNN
目录
3.3下载runtime library,developer library,code samples and the cuDNN Library
3.4安装runtime library,developer library,code samples and the cuDNN Library
1.安装NVIDIA驱动
1.1下载驱动
https://www.nvidia.com/Download/Find.aspx?lang=cn这里找对应的
418对应cuda10.1
415对应cuda10.0
以415为例
wget http://cn.download.nvidia.com/XFree86/Linux-x86_64/415.27/NVIDIA-Linux-x86_64-415.27.run其实有时候我发现浏览器直接下载也挺快的
1.2卸载原来的
sudo add-apt-repository ppa:xorg-edgers/ppa
sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt-get update
sudo apt-get remove --purge nvidia*
sudo chmod +x *.run
sudo ./NVIDIA-Linux-x86_64-384.59.run --uninstall
1.3禁用nouveau驱动
sudo vim /etc/modprobe.d/blacklist.conf在最后添加
blacklist nouveau
options nouveau modeset=0执行
sudo update-initramfs -u
使用sudo reboot重启
lsmod | grep nouveau如果没有屏幕输出，说明禁用nouveau成功
1.4禁用X-Window服务
sudo service lightdm stop然后ctrl+alt+f3进入命令行，输入账号密码
1.5命令行安装
sudo chmod +x NVIDIA-Linux-x86_64-415.27.run
sudo ./NVIDIA-Linux-x86_64-415.27.run -no-opengl-files1.6测试
#若列出GPU的信息列表，表示驱动安装成功
nvidia-smi
#若弹出设置对话框，亦表示驱动安装成功
nvidia-settings 2.安装cuda
2.1下载cuda
这里可以找到历史版本https://developer.nvidia.com/cuda-toolkit-archive
选择对应的版本下载
以10.0为例
wget https://developer.nvidia.com/compute/cuda/10.0/Prod/local_installers/cuda_10.0.130_410.48_linux其实有时候我发现浏览器直接下载也挺快的
2.2安装
chmod a+x cuda_10.0.130_410.48_linux.run
sudo ./cuda_10.0.130_410.48_linux.run --no-opengl-libs按q可以跳过文档说明
accept #同意安装
n #不安装Driver
y #安装CUDA Toolkit
<Enter> #安装到默认目录
y #创建安装目录的软链接
y #复制Samples2.3环境变量配置
sudo vim ~/.bashrc在末尾添加
export CUDA_HOME=/usr/local/cuda-10.0
export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64:$LD_LIBRARY_PATH
export PATH=/usr/local/cuda-10.0/bin:$PATH保存退出
source ~/.bashrc2.4测试
#编译并测试设备 deviceQuery：
cd /usr/local/cuda-10.0/samples/1_Utilities/deviceQuery
sudo make
./deviceQuery
#编译并测试带宽 bandwidthTest：
cd ../bandwidthTest
sudo make
./bandwidthTest如果两个都是Result = PASS，那应该是成功安装了
#查看版本
nvcc -V3.安装cudnn
3.1下载cudnn
https://developer.nvidia.com/rdp/cudnn-archive历史版本
https://developer.nvidia.com/rdp/cudnn-download最新的
要登录才能下载
看清楚版本和对应的cuda版本
以7.5为例
wget https://developer.nvidia.com/compute/machine-learning/cudnn/secure/v7.5.0.56/prod/10.0_20190219/cudnn-10.0-linux-x64-v7.5.0.56.tgz3.2安装
tar -xzvf cudnn-10.0-linux-x64-v7.5.0.56.tgz
sudo cp cuda/include/cudnn.h /usr/local/cuda/include
sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64
sudo chmod a+r /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn*3.3下载runtime library,developer library,code samples and the cuDNN Library
还是刚刚的网页
wget https://developer.nvidia.com/compute/machine-learning/cudnn/secure/v7.5.0.56/prod/10.0_20190219/Ubuntu16_04-x64/libcudnn7_7.5.0.56-1%2Bcuda10.0_amd64.deb
wget https://developer.nvidia.com/compute/machine-learning/cudnn/secure/v7.5.0.56/prod/10.0_20190219/Ubuntu16_04-x64/libcudnn7-dev_7.5.0.56-1%2Bcuda10.0_amd64.deb
wget https://developer.nvidia.com/compute/machine-learning/cudnn/secure/v7.5.0.56/prod/10.0_20190219/Ubuntu16_04-x64/libcudnn7-doc_7.5.0.56-1%2Bcuda10.0_amd64.deb3.4安装runtime library,developer library,code samples and the cuDNN Library
sudo dpkg -i libcudnn7_7.5.0.56-1+cuda10.0_amd64.deb
sudo dpkg -i libcudnn7-dev_7.5.0.56-1+cuda10.0_amd64.deb
sudo dpkg -i libcudnn7-doc_7.5.0.56-1+cuda10.0_amd64.deb3.5测试
cp -r /usr/src/cudnn_samples_v7/ $HOME
cd $HOME/cudnn_samples_v7/mnistCUDNN
make clean && make
./mnistCUDNN成功的话会输出Test passed!
更具体参看https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html