MSC
*** Need to flesh this out ***
Pytorch implementation of the Creative Adversarial Network (++, as will extend).
main.py = driver
model.py = sets up and trains the GAN
generators.py = GAN generators (DCGAN/CAN/WGAN etc.)
discriminators.py = GAN discriminators (DCGAN/CAN/WGAN etc.)
ops.py = (Pytorch) Dataset set up and some other useful stuff
utils.py = was going to store loss functions but Pytorch seems to have that covered! (Should remove)
image_check.py = Little script that checks dimensions of Wikiart dataset post image resizing
To run on Floydhub and use their Tesla K80:
floyd run --gpu --env pytorch-0.3 --data asberman/datasets/wiki256/1:/mydata "python main.py --cuda --out_folder /output"
else:
python main.py (but need to edit --dataroot and --out_folder in main.py to reflect local dir)