Hello, I’m a novice with Linux. I managed to install TensorFlow on the Anaconda environment but now I’m having a tough time getting “libcupti-dev”. This is under the Ubuntu Guide for installing Tensorflow. I worked with what guides were available but now I’m feeling glued. Has anyone ever dealt with the workaround to get libcupti-dev? Is it a file? or a library? At least have any idea how I could install obtain this for the Nvidia Cuda kit.
I installed the Cuda Toolkit kit from Nvidia and moved the files from cuDNN v5.1. to the Cuda install folder in /usr/local/cuda/ to their appropriate folders. I have tried using ‘zyppter install libcupti-dev’ but with no luck this isn’t available on their repository. Please correct me, I want to learn and get GPU support for tensorflow to work.
I recommend you install using the Docker option.
Docker enables you to run apps designed to run on other distros in an isolated environment on your openSUSE.
Skimming latest info about Docker and SUSE/openSUSE, it looks like there have been major changes and now describes only how to install the “Paid” Enterprise version of Docker.
People should bookmark my page for how to install the CE (Community “free” Edition) of Docker on both SUSE and openSUSE.
After docker has been installed, then you only need to download the image you want to run from the docker library (Looks like you’ll want the one with CUDA integration), it should be ready to “just run” without any modification. The image is likely built on a Debian or Ubuntu stub, so when you want to upgrade the image you’ll have to either use apt within the image, download a new image or modify by creating a custom Docker file, you can read the Docker specific articles and a few quick, simple exercises at https://en.opensuse.org/User:Tsu2