Ubuntu 16.04 installs Nvidia graphics driver and cuda/cudnn the pit process

Installing the deep learning framework requires the use of CUDA/CUDNN (GPU) to speed up calculations, while installing CUDA/CUDNN requires the installation of Nvidia graphics drivers first.

I encountered a driver conflict during the installation, and I had to log in two problems so that I had to reinstall the operating system again.

The information on the Internet is very messy, many are reproduced, some of the operating system version is too low, and now I write down the whole process for people who encounter the same problem reference.

Ubuntu 16.04 Default installation of third-party open source driver nouveau, install NVIDIA graphics driver first need to disable Nouveau, otherwise you will encounter conflicting issues, resulting in the inability to install Nvidia graphics card drivers.

Edit File blacklist.conf

sudo vim/etc/modprobe.d/blacklist.conf

Insert the following two lines of content in the last section of the file

Blacklist nouveau

Options Nouveau modeset=0

Update system

sudo update-initramfs-u

Reboot the system (be sure to restart)

Verify that Nouveau is disabled

Lsmod | grep nouveau

No information appears stating that the Nouveau has been disabled and the next step is to install Nvidia's graphics driver.

Press CTRL+ALT+F1 to enter text mode

sudo service LIGHTDM stop

sudo./cuda_8.0.61_375.26_linux.run

Do not install OpenGL, or even if the installation is successful, there will be a circular logon situation (at least I encountered such a situation).

Do not install the/usr/local/cuda symbolic connection (this step is optional)

sudo service LIGHTDM start

Finally copy the CUDNN header file to the/usr/local/cuda-8.0/include below, and copy the library file to/usr/local/cuda-8.0/lib64.

Sets the environment variable file. BASHRC or Profile

Export Path=/usr/local/cuda-8.0/bin: $PATH

Export ld_library_path=/usr/local/cuda-8.0/lib64: #LD_LIBRARY_PATH

Update environment variables

Souce/etc/profile

Installation is complete.

Next you can install TensorFlow, Caffe, mxnet deep learning framework.

Ubuntu 16.04 installs Nvidia graphics driver and cuda/cudnn the pit process