美文网首页
Ubuntu 16.04 NVIDIA+CUDA+cuDNN 安

Ubuntu 16.04 NVIDIA+CUDA+cuDNN 安

作者: 团不慌 | 来源:发表于2019-05-01 15:22 被阅读0次

    Ubuntu 16.04下NVIDIA驱动安装、CUDA安装、cuDNN安装文档
    参考: cuda、cudnn安装(多版本)


    #!/bin/sh
    #
    # This is an explanatory file which describes NVIDIA drivers and CUDA + cuDNN
    # Toolkit install steps.
    # 
    # Author: hwrenx (hwrenx@126.com) 2019-05-01
    
    exit 0 # should never run this script directly.
    
    # ===================================================================================================
    # CUDA Toolkit                  |   Linux x86_64 Driver Version |   Windows x86_64 Driver Version   |
    # --------------------------------------------------------------------------------------------------|
    # CUDA 10.1.105                 |           >= 418.39           |               >= 418.96           |
    # CUDA 10.0.130                 |           >= 410.48           |               >= 411.31           |
    # CUDA 9.2 (9.2.148 Update 1)   |           >= 396.37           |               >= 398.26           |
    # CUDA 9.2 (9.2.88)             |           >= 396.26           |               >= 397.44           |
    # CUDA 9.1 (9.1.85)             |           >= 390.46           |               >= 391.29           |
    # CUDA 9.0 (9.0.76)             |           >= 384.81           |               >= 385.54           |
    # CUDA 8.0 (8.0.61 GA2)         |           >= 375.26           |               >= 376.51           |
    # CUDA 8.0 (8.0.44)             |           >= 367.48           |               >= 369.30           |
    # CUDA 7.5 (7.5.16)             |           >= 352.31           |               >= 353.66           |
    # CUDA 7.0 (7.0.28)             |           >= 346.46           |               >= 347.62           |
    # ===================================================================================================
    
    # Should get NVIDIA driver via https://www.nvidia.cn/Download/index.aspx?lang=cn
    NVIDIA_RUN_FILE=NVIDIA-Linux-x86_64-384.59.run
    # CUDA version and NVIDIA driver must match
    # Could get CUDA via https://developer.nvidia.com/cuda-toolkit-archive
    
    CUDA_RUN_FILE=cuda_10.1.105_418.39_linux.run
    CUDA_VERSION=10.1
    # Corresponding cuDNN lib could be found in https://developer.nvidia.com/rdp/cudnn-archive
    CUDNN_UNPACK_PATH=.
    
    # Install needed librarys
    sudo apt-get install freeglut3-dev \
                         build-essential \
                         libx11-dev \
                         libxmu-dev \
                         ibxi-dev \
                         libgl1-mesa-glx \
                         libglu1-mesa \
                         libglu1-mesa-dev
    
    
    # Remove old nvidia drivers
    sudo apt-get remove --purge nvidia*
    # Grant permissions to the installation files
    sudo chmod +x *.run
    # Remove first in case anything remained
    sudo sh $NVIDIA_RUN_FILE --uninstall
    
    # sudo gedit /etc/modprobe.d/blacklist.conf
    # 
    # ATTENTION: should only add once
    # Disable open-source nvidia driver
    # Throw it into blacklist :)
    sudo echo "blacklist nouveau" >> /etc/modprobe.d/blacklist.conf
    sudo echo "options nouveau modeset=0" >> /etc/modprobe.d/blacklist.conf
    
    # Update
    sudo update-initramfs -u
    # Should run under cmd: Ctrl+Alt+(F1~F6)
    sudo service lightdm stop
    
    # Install driver, must disable openGL under ubuntu 16+
    # Or there may have loop login appearance
    # Choosing all default option is ok
    sudo sh $NVIDIA_RUN_FILE –no-opengl-files   # --no-x-check --no-nouveau-check --disable-nouveau
    
    # Try print driver log to check whether the installation is successful
    nvidia-smi
    
    # Restart the graphical interface
    # Ctrl+Alt+F7
    sudo service lightdm restart
    # Reboot
    sudo reboot
    
    # Install CUDA
    sudo sh $CUDA_RUN_FILE --no-opengl-libs
    
    # ATTENTION: should only add once
    echo "export PATH=/usr/local/cuda/bin:$PATH" >>~/.bashrc
    echo "export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH" >>~/.bashrc
    echo "export CUDA_HOME=/usr/local/cuda" >>~/.bashrc
    
    # Link cuda files considering multi-version condition
    sudo ln -s /usr/local/cuda-$CUDA_VERSION /usr/local/cuda
    
    # Validate
    source ~/.bashrc
    
    # Try CUDA
    nvcc --version
    
    # Copy cuDNN files into CUDA
    cd $CUDNN_UNPACK_PATH
    sudo cp cuda/include/cudnn.h /usr/local/cuda-$CUDA_VERSION/include/ 
    sudo cp cuda/lib64/libcudnn*  /usr/local/cuda-$CUDA_VERSION/lib64
    sudo chmod a+r /usr/local/cuda-$CUDA_VERSION/include/cudnn.h /usr/local/cuda-$CUDA_VERSION/lib64/libcudnn*
    
    
    

    相关文章

      网友评论

          本文标题:Ubuntu 16.04 NVIDIA+CUDA+cuDNN 安

          本文链接:https://www.haomeiwen.com/subject/yqfwnqtx.html