ExoGAN Guide

ExoGAN is written in Tensorflow (v. 1.8) with additional standard python packages.

#Installing Tensorflow

Tensorflow is easy to install on most systems though we do find GPU support to be difficult to install and highly architecture dependent. Hene we will not provide a general installation manual here but refer you to the tensorflow website.

https://www.tensorflow.org/install/

If possible, we strongly suggest to run tensorflow in docker container (avaialable on the tensorflow website). If your cluster supports containers, it is a much easier way to port your code and handle difficult GPU dependencies.

We have also created a singularity container for use on clusters supporting singularity (https://singularity.lbl.gov). For the singularity image description, please see here.

#Installing python packages

We recommend using anaconda as python distribution but python packages can easily be installed using pip

pip install scipy numpy matplotlib argparse