??最近老板突然要說(shuō)做項(xiàng)目，雙手一揮就申請(qǐng)了張顯卡，因此記錄下這篇文檔；

系統(tǒng)?/?Ubunto16.04
顯卡?/?Nvidia GTX 1070ti

• NVIDIA顯卡驅(qū)動(dòng)

  • 安裝準(zhǔn)備

    • 屏蔽nouveau開(kāi)源驅(qū)動(dòng)

    touch /etc/modprobe.d/blacklist-nouveau.conf 
    echo "blacklist nouveau" >>blacklist-nouveau.conf 
    echo "options nouveau modeset = 0" >>blacklist-nouveau.conf
    

    • 更新前可以去blacklist-nouveau.conf查看命令是否添加成功，之后執(zhí)行更新：

    sudo update-initramfs -u
    

    • 去Nvidia官網(wǎng)下載和顯卡對(duì)應(yīng)的驅(qū)動(dòng)，我的是GTX1070ti，對(duì)應(yīng)的最新的驅(qū)動(dòng)是NVIDIA-Linux-x86_64-390.48.run

  • 安裝NVIDIA顯卡驅(qū)動(dòng)：

    • 進(jìn)入字符界面Ctrl+alt+F1之后，輸入同戶名和密碼，登陸成功后執(zhí)行：

    sudo service lightdm stop
    

    • 安裝:其中–no-opengl-files很重要，不然安裝后重啟會(huì)出現(xiàn)循環(huán)登錄的問(wèn)題。

     sudo chmod 777 NVIDIA-Linux-x86_64-390.48.run   //執(zhí)行權(quán)限
     sudo sh NVIDIA-Linux-x86_64-390.48.run –no-opengl-files   //執(zhí)行
     sudo service lightdm start
     sudo reboot
    

    • 重啟如果能夠順利登錄，恭喜，之后測(cè)試是否安裝成功：

    nvidia-smi
    

    打印出gpu相關(guān)信息表示安裝成功。

  • 安裝CUDA-8.0

    • 安裝依賴：
      • 設(shè)置源：

      # deb cdrom:[Ubuntu 16.04 LTS _Xenial Xerus_ - Release amd64 (20160420.1)]/ xenial main restricted
      deb-src http://archive.ubuntu.com/ubuntu xenial main restricted #Added by software-properties
      deb http://mirrors.aliyun.com/ubuntu/ xenial main restricted
      deb-src http://mirrors.aliyun.com/ubuntu/ xenial main restricted multiverse universe #Added by software-properties
      deb http://mirrors.aliyun.com/ubuntu/ xenial-updates main restricted
      deb-src http://mirrors.aliyun.com/ubuntu/ xenial-updates main restricted multiverse universe #Added by software-properties
      deb http://mirrors.aliyun.com/ubuntu/ xenial universe
      deb http://mirrors.aliyun.com/ubuntu/ xenial-updates universe
      deb http://mirrors.aliyun.com/ubuntu/ xenial multiverse
      deb http://mirrors.aliyun.com/ubuntu/ xenial-updates multiverse
      deb http://mirrors.aliyun.com/ubuntu/ xenial-backports main restricted universe multiverse
      deb-src http://mirrors.aliyun.com/ubuntu/ xenial-backports main restricted universe multiverse #Added by software-properties
      deb http://archive.canonical.com/ubuntu xenial partner
      deb-src http://archive.canonical.com/ubuntu xenial partner
      deb http://mirrors.aliyun.com/ubuntu/ xenial-security main restricted
      deb-src http://mirrors.aliyun.com/ubuntu/ xenial-security main restricted multiverse universe #Added by software-properties
      deb http://mirrors.aliyun.com/ubuntu/ xenial-security universe
      deb http://mirrors.aliyun.com/ubuntu/ xenial-security multiverse
      

      將aliyun的源添加到/etc/apt/source.list中;
    • 安裝相關(guān)依賴庫(kù)：

      sudo apt-get install freeglut3-dev build-essential libx11-dev
      sudo apt-get install libxmu-dev libxi-dev libgl1-mesa-glx libglu1-mesa
      sudo apt-get install libglu1-mesa-dev
      ```
    

  • 安裝：ubunto16系統(tǒng)默認(rèn)的gcc-5.4.0就支持cuda-8.0，我的cuda-runfile文件是cuda_8.0.61_375.26_linux-run

     sudo sh cuda_8.0.44_linux.run --no-opengl-libs
    

    這里沒(méi)有安裝opengl，不會(huì)出現(xiàn)循環(huán)登錄的bug；

    • 添加環(huán)境變量

    vim ~/.bashrc
    export PATH=/usr/local/cuda-8.0/bin:$PATH
    export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:$LD_LIBRARY_PATH
    sudo vim /etc/profile
    export CUDA_HOME=/usr/local/cuda-8.0
    

    • 設(shè)置動(dòng)態(tài)鏈接庫(kù)

    sudo vim /etc/profile
    

    寫(xiě)入

    export PATH = /usr/local/cuda/bin:$PATH
    

    創(chuàng)建cuda.conf文件

    sudo gedit /etc/ld.so.conf.d/cuda.conf
    

    添加以下路徑

    /usr/local/cuda/lib64
    

    執(zhí)行鏈接生效

    sudo ldconfig
    sudo reboot
    

    • 測(cè)試cuda是否安裝成功

    cd /usr/local/cuda-8.0/samples/1_Utilities/deviceQuery 
    sudo make 
    ./deviceQuery
    

    得到以下結(jié)果表示安裝成功。

  • 安裝cuDNN-5.1

    • Cuda8.0對(duì)應(yīng)的cnDNN版本是5.1，去官網(wǎng)注冊(cè)下載；
    • 下載之后解壓，將cuDNN里的文件copy到CUDA目錄；

    sudo cp cudnn.h /usr/local/cuda/include/ 
    sudo cp lib* /usr/local/cuda/lib64/
    cd /usr/local/cuda/lib64/ 
    sudo rm -rf libcudnn.so libcudnn.so.5
    sudo ln -s libcudnn.so.5.1.5 libcudnn.so.5 
    sudo ln -s libcudnn.so.5 libcudnn.so 
    

  • 安裝python

    • 安裝setuptools依賴的zlib庫(kù)；

    download:http://www.zlib.net/
    ./configure --prefix=/usr/local/zlib/
    make
    make install
    

    添加鏈接；

    //將--prefix目錄添加到zlib.conf中
    sudo vim /etc/ld.so.conf.d/zlib.conf
    ldconfig
    

    • 安裝setuptools；

    download:https://pypi.org/project/setuptools/
    sudo python setup.py install
    

    • 安裝pip；

    sudo python setup.py install
    

  • 安裝tensorflow

    pip安裝：

    //gpu-python2
    sudo pip install --ignore-installed --upgrade https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.2.0-cp27-none-linux_x86_64.whl
    //cpu-python2
    sudo pip install --ignore-installed --upgrade https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.2.0-cp27-none-linux_x86_64.whl
    //gpu-python3
    sudo pip install --ignore-installed --upgrade https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-1.2.0-cp34-cp34m-linux_x86_64.whl
    //cpu-python3
    sudo pip install --ignore-installed --upgrade https://storage.googleapis.com/tensorflow/linux/gpu/tensorflow_gpu-1.2.0-cp34-cp34m-linux_x86_64.whl
    

  • 總結(jié)

    這樣環(huán)境就搭好了，可以愉快的燒GPU啦～