Deep Learning Workbench Installation Notes

CUDA 8.0

nVidia显卡驱动

更新一下

$ sudo apt-get update
$ sudo apt-get upgrade

添加显卡驱动的PPA

$ sudo add-apt-repository ppa:graphics-drivers/ppa

然后显示信息

在这里回车确认

下面就可以开始安装显卡驱动了，下载速度有点慢。

$ sudo apt-get update
$ sudo apt-get install nvidia-375

驱动安装完成一定要重启系统

测试

使用nvidia-smi检查显卡状况。

• Download from the official website:
  • Select Linux > x86_64 > Uuntu > 14.4 > deb(local), then download the according package (about 1.9 GB).
• Installation Instructions:
  • Simply use my automation script in my github.

  • Answer some questions as follows

    • Do you accept the previously read EULA? : accept
    • Install NVIDIA Accelerated Graphics Driver for ~~ : n
    • Install the CUDA 7.5 Toolkit? : y
    • Enter Toolkit Location : (blank)
    • Do you want to install a symbolic link at ~~ : y
    • Install the CUDA 7.5 Samples? : y
    • Enter CUDA Samples Location : (blank)

cuDNN 6

• One needs to register with an email to download.
• $ sudo tar -xvf cudnn-8.0-linux-x64-v6.0-rc.tgz -C /usr/local

tensorflow

• Install Anaconda
• Under a virtual envrionment, use pip install

  • $ pip install tensorflow
    
    or when you have GPU
    
    $ pip install tensorflow-gpu