step1.安装cuda+cudnn:
0.检查型号:
lspci | grep -i nvidia
1.检查自己显卡能安装什么版本的cuda:CUDA GPUs - Compute Capability | NVIDIA Developer
知道自己显卡型号并查询下载:官方驱动 | NVIDIA
2.安装cuda:CUDA Toolkit Archive | NVIDIA Developer
3.安装 cudnn:cuDNN Download | NVIDIA Developer
4.安装pytorch: Start Locally | PyTorch
Cuda安装
ubuntu20.04神舟安装步骤:
1.sudo gedit /etc/modprobe.d/blacklist.conf
在最后写入
blacklist nouveau
options nouveau modeset=0然后终端输入
sudo update-initramfs -u #应用更改
2.安装显卡(20.04自己会安装
ps.18.04下载对应版本驱动
sudo chmod a+x NVIDIA-Linux-x86_64-465.31.run sudo ./NVIDIA-Linux-x86_64-465.31.run -no-opengl-files
3.输入nvidia-smi检查自己cuda支持的最高版本,并根据以后要使用的pytorch和tensorflow选择自己需要的cuda版本。这里cuda最高支持12.1。但是我们选择下载cuda11.6版本
4.cuda11.6下载链接:CUDA Toolkit 11.6 Downloads | NVIDIA Developer
4-1.安装一些相互依赖的库文件,在终端运行下面指令:
sudo apt-get install freeglut3-dev build-essential libx11-dev libxmu-dev libxi-dev libgl1-mesa-glx libglu1-mesa libglu1-mesa-dev
5.终端依次输入第一条命令,并在下载目录下打开终端输入第二条命令
$wget https://developer.download.nvidia.com/compute/cuda/11.6.0/local_installers/cuda_11.6.0_510.39.01_linux.run $sudo sh cuda_11.6.0_510.39.01_linux.run
例如,这里下载在桌面,则直接在终端输入
sudo sh cuda_11.6.0_510.39.01_linux.run
6.弹框选择continue,然后输入accept
等待后显示下图,取消第一个driver,然后在Install回车
出现下图,选择yes,并显示如下。
7.配置cuda环境,根据自己的cuda版本(可以去文件路径下查看一下对应的是否正确
sudo gedit ~/.bashrc
export PATH=/usr/local/cuda-11.6/bin${PATH:+:${PATH}} export LD_LIBRARY_PATH=/usr/local/cuda-11.6/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
source ~/.bashrc
8. CUDA测试,终端输入nvcc -V。
可以看到输出Cuda compilation tools, release 11.6, V11.6.55
cuDNN安装
1.去官网cuDNN Archive | NVIDIA Developer。选择对应版本的cuDNN,因为我的CUDA是11.6版本的,选择第二个。
这个也选第二个,tar版本的。(需要注册nvidia账号,邮箱注册好之后可以直接下载
2.从‘下载’文件夹下把东西移动到自己新建的‘cudnn’文件夹中,然后依次输入(改成自己文件夹的名字,不要直接复制!!
$xz -d cudnn-linux-x86_64-8.9.0.131_cuda11-archive.tar.xz $tar xf cudnn-linux-x86_64-8.9.0.131_cuda11-archive.tar
3.终端依次输入(这个可以直接复制
sudo cp cudnn-*-archive/include/cudnn*.h /usr/local/cuda/include sudo cp -P cudnn-*-archive/lib/libcudnn* /usr/local/cuda/lib64 sudo chmod a+r /usr/local/cuda/include/cudnn*.h /usr/local/cuda/lib64/libcudnn*
4.终端测试cudnn是否下载成功
cat /usr/local/cuda-11.6/include/cudnn_version.h | grep CUDNN_MAJOR -A 2
Win10安装cuda+cudnn参考:
Windows10安装Pytorch的完整安装教程_windows 安装pytorch_知行达摩院的博客-CSDN博客
win10的pytorch安装(初学者经验总结)_win10 pytorch_三思为上策的博客-CSDN博客
三十分钟在Windows10安装Anaconda+Pytorch+Cuda
【Windows】Win10的anaconda安装+Anaconda更换镜像源+conda常用指令_win10 anaconda换源
MX450英伟达链接:GeForce MX450 笔记本电脑显卡 | NVIDIA
GeForce MX450 笔记本电脑显卡 | NVIDIAGeForce MX450 笔记本电脑显卡 | NVIDIA
Ubuntu20.04
Q:nvidia-smi No devices were found
ubuntu-drivers devices
sudo apt install nvidia-driver-525-server
reboot
===========
= Summary =
===========Driver: Not Selected
Toolkit: Installed in /usr/local/cuda-11.7/Please make sure that
- PATH includes /usr/local/cuda-11.7/bin
- LD_LIBRARY_PATH includes /usr/local/cuda-11.7/lib64, or, add /usr/local/cuda-11.7/lib64 to /etc/ld.so.conf and run ldconfig as rootTo uninstall the CUDA Toolkit, run cuda-uninstaller in /usr/local/cuda-11.7/bin
***WARNING: Incomplete installation! This installation did not install the CUDA Driver. A driver of version at least 515.00 is required for CUDA 11.7 functionality to work.
To install the driver using this installer, run the following command, replacing <CudaInstaller> with the name of this run file:
sudo <CudaInstaller>.run --silent --driverLogfile is /var/log/cuda-installer.log