https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html#title-new-features
参考https://phantomvk.github.io/2019/06/29/Ubuntu_install_nVidia_Driver/
添加PPA,过程中回车确认询问的信息
sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt update
ubuntu-drivers devices
根据下列结果,这里 nvidia-driver-465 为推荐驱动安装版本(但cuda11.3要求465以上)
== /sys/devices/pci0000:64/0000:64:00.0/0000:65:00.0 ==
modalias : pci:v000010DEd00002204sv00001028sd00003880bc03sc00i00
vendor : NVIDIA Corporation
driver : nvidia-driver-460 - distro non-free
driver : nvidia-driver-460-server - distro non-free
driver : nvidia-driver-465 - distro non-free recommended
driver : xserver-xorg-video-nouveau - distro free builtin
直接安装最新驱动,安装完成后重启电脑
sudo apt install nvidia-driver-465
也可以用sudo ubuntu-drivers autoinstall 会自动安装推荐的版本(一般是最高的版本)重启后用命令 nvidia-smi 检查显卡是否被正常识别:显卡P106-100,显存6G,驱动430.26,CUDA10.2(应显示驱动为440)
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 430.26 Driver Version: 430.26 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 P106-100 Off | 00000000:01:00.0 Off | N/A |
| 39% 40C P0 27W / 120W | 0MiB / 6080MiB | 4% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+
如果出现NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running. 需要手动关闭BIOS的安全模式。
如果安装 nvidia-driver-410 或以上版本提示 packages 无法安装,请执行以下步骤:
sudo apt-add-repository -r ppa:graphics-drivers/ppa
更新 apt
sudo apt update
移除 NVIDIA 显卡驱动文件
sudo apt remove nvidia*
执行自动清理
sudo apt autoremove
然后重新回到本文初步骤重新安装
根据配置选择cuda的版本 https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_64&Distribution=Ubuntu&target_version=18.04&target_type=runfile_local
wget wget https://developer.download.nvidia.com/compute/cuda/11.3.1/local_installers/cuda_11.3.1_465.19.01_linux.run
sudo sh cuda_10.2.89_440.33.01_linux.run
参考 https://blog.csdn.net/u013084111/article/details/104167056
注册并下载以下3个文件
cuDNN Runtime Library for Ubuntu18.04 (Deb)
cuDNN Developer Library for Ubuntu18.04 (Deb)
cuDNN Code Samples and User Guide for Ubuntu18.04 (Deb)
要下载deb文件,不要下载tgz文件!(亲测tgz易错)
网址
https://developer.nvidia.com/rdp/cudnn-download
进入到下载文件所在目录
$ sudo dpkg -i libcudnn7_7.6.5.32-1+cuda10.2_amd64.deb
$ sudo dpkg -i libcudnn7-dev_7.6.5.32-1+cuda10.2_amd64.deb
$ sudo dpkg -i libcudnn7-doc_7.6.5.32-1+cuda10.2_amd64.deb
检测cuDNN是否安装成功:
运行/usr/src/cudnn_samples_v7 中的mnistCUDNN sample
$ cd /usr/src/cudnn_samples_v7
$ cp -r /usr/src/cudnn_samples_v7/ $HOME
$ cd $HOME/cudnn_samples_v7/mnistCUDNN
$ make clean && make
$ ./mnistCUDNN
如果出现 Test passed! 则说明安装成功
如果编译时出现fatal error: FreeImage.h: No such file or directory错误,参考https://blog.csdn.net/xhw205/article/details/116297555
cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2
已经不能用了,因为cudnn.h文件里内容修改了,不再存放版本信息。cat /usr/local/cuda/include/cudnn.h | grep cudnn
查看,发现里面导入了cudnn_version.h文件,版本信息就存放在这个文件里
使用find命令找到对应的文件就可以了
find / -name cudnn_version.h 2>&1 | grep -v "Permission denied"
cat /usr/include/cudnn_version.h | grep CUDNN_MAJOR -A 2