Ishaan Gulrajani

Thanks for the PR :) Re. the Inception Score changes, are those bugs also present in the canonical implementation (https://github.com/openai/improved-gan/blob/master/inception_score/model.py)? If so then I'm hesitant to change because enough people...

The CIFAR10 model which got the best results was a resnet, but not 101-resnet. We trained 101-resnet on bedrooms (see gan_64x64.py) to demonstrate that it worked, but we never expected...

It seems like you're trying to run the model without a GPU. Is there any reason you want to do this? I think it would be too slow to be...

sorry, the batchnorm implementation is a mess. i think this is what you need to do: 1. train with is_training=True (update_moving_stats and stats_iter can be anything) 2. after training, run...

Those don't typically apply. You in fact want D loss to be a large negative value within the first ~1000 iters, and then want it to increase steadily (ie move...

I can put something together, but we ran experiments with this and found it didn't perform nearly as well as 2-level. Do you want it anyway?

Just updated the code with a model that should work ('64px_big_onelevel'), but I haven't actually tested it. Let me know how it goes if you do.

Interesting! SOTA in what sense, sample quality? I'd say LSUN bedrooms fits the description you mentioned (plenty of data, not terribly complex images) and we tried to demonstrate pretty good...

Can you point to the sampling code you're talking about? The code that gets executed should be lines 855-902.