Skip to content
No description, website, or topics provided.
Jupyter Notebook Python
Branch: master
Clone or download
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
networks Fix save_every_step bug Aug 11, 2018
slides Add SLAC slides Oct 4, 2017
.gitignore First commit Jun 22, 2017
README.md Update Feb 9, 2018

README.md

cosmoGAN

This code is to accompany "Creating Virtual Universes Using Generative Adversarial Networks" manuscript arXiv:1706.02390. The architecture is an implementation of the DCGAN architecture (arXiv:1511.06434).


How to train:

git clone git@github.com:MustafaMustafa/cosmoGAN.git
cd cosmoGAN/networks
mkdir data
wget http://portal.nersc.gov/project/dasrepo/cosmogan/cosmogan_maps_256_8k_1.npy
cd ../

That will download sample data (8k maps) for testing. You can download more data from here. All of this data has been generated using our GAN and can be used to train your own. Original data is available upon request from the authors.

To run:

python run_dcgan.py

Load pre-trained weights:

First download the weights:

cd cosmoGAN/networks
wget http://portal.nersc.gov/project/dasrepo/cosmogan/cosmoGAN_pretrained_weights.tar
tar -xvf cosmoGAN_pretrained_weights.tar

Then take a look at networks/load_and_use_pretrained_weights.ipynb notebook for how to run.

You can’t perform that action at this time.