zoukankan      html  css  js  c++  java
  • ubuntu安装CUDA及cuDNN

    0 写在前面

    安装环境:ubuntu18.04以及GTX1050Ti笔记本

    为什么要安装CUDA?
    参考百科,CUDA是英伟达推出的集成技术,通过该技术可利用GeForce 8 以后的GPU或者较新的Quadro GPU进行计算。例如典型的tensorflow-GPU和pyCUDA安装之前都要先安装CUDA。

    1 安装N卡驱动

    安装ubuntu系统之后自带开源NVIDIA Nouveau驱动,但是很容易出现ubuntu18双系统安装后登陆重启卡死问题。安装N卡驱动(即CUDA的硬件支持)之前必须先禁用这个驱动。
    终端输入以下命令没有返回结果说明禁用成功。

    lsmod | grep nouveau
    

    然后终端输入如下可以查看推荐的驱动版本:

    ubuntu-drivers devices
    

    笔者显示结果如下,说明nvidia-driver-435是推荐的

    jj@jj-u:~/Downloads$ ubuntu-drivers devices
    == /sys/devices/pci0000:00/0000:00:01.0/0000:01:00.0 ==
    modalias : pci:v000010DEd00001C8Csv0000103Csd0000838Fbc03sc00i00
    vendor   : NVIDIA Corporation
    model    : GP107M [GeForce GTX 1050 Ti Mobile]
    driver   : nvidia-driver-435 - distro non-free recommended
    driver   : nvidia-driver-430 - distro non-free
    driver   : nvidia-driver-390 - distro non-free
    driver   : nvidia-driver-410 - third-party free
    driver   : xserver-xorg-video-nouveau - distro free builtin
    

    此时终端输入sudo ubuntu-drivers autoinstall即可自动安装,或者输入sudo apt install nvidia-driver-435安装,然后重启系统即可。

    安装过程中可能会跳出让你进入什么Key界面,这是因为安全模式安装第三方驱动需要写入key,具体遇到可百度,最简单粗暴的方法就是进入BIOS关闭安全模式启动,然后再安装n卡驱动。

    重启后终端输入nvidia-smi,结果如下:

    jj@jj-u:~/Downloads$ nvidia-smi
    Tue Jan  7 12:31:39 2020       
    +-----------------------------------------------------------------------------+
    | NVIDIA-SMI 435.21       Driver Version: 435.21       CUDA Version: 10.1     |
    |-------------------------------+----------------------+----------------------+
    | GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
    | Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
    |===============================+======================+======================|
    |   0  GeForce GTX 105...  Off  | 00000000:01:00.0  On |                  N/A |
    | N/A   42C    P0    N/A /  N/A |    523MiB /  4040MiB |     18%      Default |
    +-------------------------------+----------------------+----------------------+
                                                                                   
    +-----------------------------------------------------------------------------+
    | Processes:                                                       GPU Memory |
    |  GPU       PID   Type   Process name                             Usage      |
    |=============================================================================|
    |    0      1491      G   /usr/lib/xorg/Xorg                           210MiB |
    |    0      1658      G   /usr/bin/gnome-shell                         146MiB |
    |    0      2808      G   /proc/self/exe                                12MiB |
    |    0      7951      G   ...quest-channel-token=9976497191416364032   142MiB |
    |    0      8743      G   /opt/teamviewer/tv_bin/TeamViewer              9MiB |
    +-----------------------------------------------------------------------------+
    

    可以看到该驱动支持的最高CUDA版本是10.1,刻意提及最高版本是因为某些比如tensorflow-gpu 1.x版本或最高支持CUDA9.1,但是驱动支持向下兼容所以可以安装CUDA9.1,但最好版本别太老。
    然后终端输入nvidia-settings出现图形设置界面说明到此N卡驱动安装成功。

    2 安装CUDA

    2.1 下载安装CUDA

    点击CUDA各版本链接,由于tensorflow 1.x版本可能不兼容CUDA10.1,所以选择CUDA10,按提示下载了2G左右的runfile,然后按提示终端输入sudo sh cuda_10.1.105_418.39_linux.run,然后一直按F键读条到100%,然后根据终端提示输入即可,NVIDIA Accelerated Graphics Driver由于上一步安装过驱动就否了,

    Do you accept the previously read EULA?
    accept/decline/quit: accept          
    
    Install NVIDIA Accelerated Graphics Driver for Linux-x86_64 410.48?
    (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/jj ]: 
    
    Installing the CUDA Toolkit in /usr/local/cuda-10.0 ...
    

    2.2 配置环境变量

    终端输入:

    cd ~
    sudo gedit .bashrc
    

    打开文档后末尾加入以下信息:

    export LD_LIBRARY_PATH=/usr/local/cuda/lib64:/usr/local/cuda/extras/CPUTI/lib64
    export CUDA_HOME=/usr/local/cuda/bin
    export PATH=$PATH:$LD_LIBRARY_PATH:$CUDA_HOME
    

    然后source .bashrc或者重启XD即可,终端输入nvcc -V可检查是否安装成功,结果如下:

    nvcc: NVIDIA (R) Cuda compiler driver
    Copyright (c) 2005-2018 NVIDIA Corporation
    Built on Sat_Aug_25_21:08:01_CDT_2018
    Cuda compilation tools, release 10.0, V10.0.130
    

    或者终端输入:

    cd ~/NVIDIA_CUDA-10.0_Samples/1_Utilities/bandwidthTest
    make
    ./bandwidthTest
    

    返回Result = PASS代表cuda安装成功。

    3 安装cuDNN

    cuDNN是用于深度神经网络的GPU加速库,如果CUDA是工作台,那么cuDNN就是上面的螺丝刀等工具。
    下载安装参考CSDN。

  • 相关阅读:
    HDU 1058 Humble Numbers
    HDU 1160 FatMouse's Speed
    HDU 1087 Super Jumping! Jumping! Jumping!
    HDU 1003 Max Sum
    HDU 1297 Children’s Queue
    UVA1584环状序列 Circular Sequence
    UVA442 矩阵链乘 Matrix Chain Multiplication
    DjangoModels修改后出现You are trying to add a non-nullable field 'download' to book without a default; we can't do that (the database needs something to populate existing rows). Please select a fix:
    opencv做的简单播放器
    c++文件流输入输出
  • 原文地址:https://www.cnblogs.com/hzcya1995/p/13281722.html
Copyright © 2011-2022 走看看