Skip to content

ethancaballero/ENAS_GAN

Repository files navigation

ENAS_GAN

collaboration between: Ethan Caballero & Aran Komatsuzaki

This is an attempt to use ENAS (Efficient Neural Architecture Search) to find optimal GAN architectures. Unfortunately, ENAS seems to be unable to find useful GAN architectures.

First download CIFAR-10 python version from https://www.cs.toronto.edu/~kriz/cifar.html to obtain folder cifar-10-batches-py/

to run: python3 train.py --save True --incept-start-epoch 10

to run ppo: python3 train.py --save True --incept-start-epoch 10 --ppo True

Make sure to change the names of saved models (or back them up somewhere) before running new models that will be saved, because the new saved models will overwrite the old saved models if old saved models have default names.

Dependencies

pytorch v0.2 or v0.3

About

No description or website provided.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages