Before You Start
Verify GPU and stuff.
$ lspci | grep -i nvidia $ uname -m && cat /etc/*release $ gcc --version
Make sure you download the correct version. I am using Ubuntu 16, so I need CUDA 8. I made the mistake of downloading CUDA 7.5, which is only compatible with Ubuntu 14 and 15.
Dowload from NVIDIA: https://developer.nvidia.com/cuda-toolkit
As of the time of this post, the installation instruction on the download page has a typo in it. The file name has a “-deb” instead of “.deb“.
$ sudo dpkg -i cuda-repo-ubuntu1604-8-0-rc_8.0.27-1_amd64.deb $ sudo apt-get update $ sudo apt-get install cuda
After installing we will need to add the path.
$ export PATH=/usr/local/cuda-8.0/bin:$PATH $ export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:$LD_LIBRARY_PATH
Verify the driver.
$ cat /proc/driver/nvidia/version NVRM version: NVIDIA UNIX x86_64 Kernel Module 361.62 Tue May 24 20:21:31 PDT 2016 GCC version: gcc version 5.4.0 20160609 (Ubuntu 5.4.0-6ubuntu1~16.04.1)
$ nvcc -V The program 'nvcc' is currently not installed. You can install it by typing: sudo apt install nvidia-cuda-toolkit $ sudo apt install nvidia-cuda-toolkit
If the compiler was not install, install it now. The result should look something like this.
$ nvcc -V nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2015 NVIDIA Corporation Built on Tue_Aug_11_14:27:32_CDT_2015 Cuda compilation tools, release 7.5, V7.5.17
At this point, you’re supposed to run the samples in order to test if everything is working properly. I was never able to install the samples, but I was able to get Tensorflow and Torch to run using the GPU. So, in the end, everything worked out fine.
This is easy. All you have to do is uncompressed a folder and then copy and paste the files to the correct locations.
$ sudo cp include/cudnn.h /usr/local/cuda/include $ sudo cp lib64/libcudnn* /usr/local/cuda/lib64 $ sudo chmod a+r /usr/local/cuda/lib64/libcudnn*
links and tutorials: