- 查看显卡驱动
nvidia-smi
- 查看cuda版本
cat /usr/local/cuda/version.txt nvcc -V
- 查看cudnn版本
cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2
- 下载驱动
- https://www.nvidia.cn/geforce/drivers
- 下载NVIDIA-Linux-x86_64-440.100.run文件，大小为144MB
- 安装
# 1.禁用nouveau sudo gedit /etc/modprobe.d/blacklist.conf # 打开文件，在最后添加如下两行 blacklist nouveau options nouveau modeset=0 # 2.更新系统修改 sudo update-initramfs -u # 重启电脑 # 3.验证nouveau已禁用 lsmod | grep nouveau # 无输出则表示已禁用 # 4.按ctrl+alt+f1进入命令行界面，此时需要login（电脑账户名称），password（密码） # 5.关闭图形界面 sudo service lightdm stop # 6.卸载系统中存在的驱动 sudo apt-get remove nvidia-* # 7.运行.run文件 sudo chmod a+x NVIDIA-Linux-x86_64-440.100.run sudo ./NVIDIA-Linux-x86_64-440.100.run -no-x-check -no-nouveau-check -no-opengl-files # 在弹出的提示窗口中均保持默认选择 # 8.重启图形界面 sudo service lightdm start
- 验证安装成功
nvidia-smi
- 安装前提
- 已安装显卡驱动
- 下载驱动
- https://developer.nvidia.com/cuda-toolkit-archive
- 下载cuda_10.0.130_410.48_linux.run文件[CUDA Toolkit 10.0 Archive -> Linux -> x86_64 -> Ubuntu -> 16.04 -> runfile(local)]，大小为2.0GB
sudo sh cuda_10.0.130_410.48_linux.run
- 如果/tmp显示空间不足，则执行
sudo mkdir /opt/tmp sudo sh cuda_10.0.130_410.48_linux.run --tmpdir=/opt/tmp/
- 安装
# Do you accept the previously read EULA? # accept/decline/quit: accept # Install NVIDIA Accelerated Graphics Driver for Linux-x86_64 361.62? # (y)es/(n)o/(q)uit: n # Install the CUDA 10.0 Toolkit? # (y)es/(n)o/(q)uit: y # Enter Toolkit Location # [ default is /usr/local/cuda-10.0 ]: # Do you want to install a symbolic link at /usr/local/cuda? # (y)es/(n)o/(q)uit: y # Install the CUDA 10.0 Samples? # (y)es/(n)o/(q)uit: y # Enter CUDA Samples Location # [ default is /home/lishangjie ]:
- 安装结束后，在~/.bashrc最后添加
export PATH=/usr/local/cuda-10.0/bin:$PATH export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
- 验证安装成功
cd /usr/local/cuda-10.0/samples/1_Utilities/deviceQuery sudo make ./deviceQuery # 不报错即证明安装成功
- 查看cuda版本（CUDA Version 10.0.130）
cat /usr/local/cuda-10.0/version.txt
- 多版本cuda切换
# 删除之前的cuda软链接，并根据版本需要重新创建 sudo rm -rf /usr/local/cuda sudo ln -s /usr/local/cuda-10.0 /usr/local/cuda # 在~/.bashrc中使用cuda软链接（无需指明版本号）
- 安装前提
- 已安装cuda
- 下载驱动（首次登录需注册并填问卷）
- https://developer.nvidia.com/rdp/cudnn-archive
- 下载cudnn-10.0-linux-x64-v7.6.5.32.tgz压缩包[Download cuDNN v7.6.5 (November 5th, 2019), for CUDA 10.0 -> cuDNN Library for Linux]，大小为486MB
- 解压后生成cuda文件夹
sudo cp cuda/include/cudnn.h /usr/local/cuda/include sudo cp cuda/lib64/*.* /usr/local/cuda/lib64 sudo chmod a+r /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn*
- 查看cudnn版本（CUDNN_VERSION 7.6.5）
cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2
-
设置软件和更新
- 在软件和更新中，选择阿里的服务器（http://mirrors.aliyun.com/ubuntu）
-
若重新载入时出错，修改/etc/apt/sources.list
# 删除文档原内容，并添加 deb http://mirrors.aliyun.com/ubuntu/ bionic main restricted universe deb http://mirrors.aliyun.com/ubuntu/ bionic-security main restricted universe deb http://mirrors.aliyun.com/ubuntu/ bionic-updates main restricted universe deb http://mirrors.aliyun.com/ubuntu/ bionic-proposed main restricted universe deb http://mirrors.aliyun.com/ubuntu/ bionic-backports main restricted universe
