Add GAN models with inference networks about torchgan HOT 3 OPEN

It looks like we might have to bring changes into the training loop itself and generalise it further to account for inference model based GANs.
@yatindandi its best if you write tutorials for all the three models first and then make appropriate changes to the training loop.

from torchgan.

@yatindandi I believe we should put the Adversarial Variational Bayes in the model zoo instead of this repo. It is a great paper but as per the original discussions with @Aniket1998 we were going to restrict core implementations to papers with over 250 citations.

I am happy to hear you views on this.

from torchgan.

Also some starter work on this is in the ap/hybrid branch

from torchgan.