IntroVAE

IntroVAE copied to clipboard

I guess there is typo in your paper "distinguishable"

3

 I think this should be indistinguishable. isn't it? I was little bit confused by this typo.

junseokoh1

https://github.com/hhb072/IntroVAE/blob/c8ce5d291fe8e66189d70b3ceddc2eb2266d3742/main.py#L191 It seems more natural of the converse case, i.e., relu(...).mean().

GloryyrolG

This work cannot be true

1

In your paper and your code, the KL term (regularization term) is exactly negative KL divergence of the approximate posterior with the prior, while in VAE it should be positive....

ycxl2020

Would like to knwo if you will release the pretrained model, if will help a lot. Thank you!

SystemErrorWang

How to understand the para `weight_neg, weight_rec, weight_kl`? What is the relationship between them and `alpha, beta` in Introvae paper?

leoliu37

Crashes on 1-GPU with log access error

1

When I run IntroVAE on 1 GPU (to test how it works on my [anime faces](https://www.gwern.net/Faces#introvae)), I get indexing/scalar errors from PyTorch (`TypeError: only integer scalar arrays can be converted...

gwern

WHY add before bn?

1

on https://github.com/hhb072/IntroVAE/blob/master/networks.py line 40 `output = self.relu2(self.bn2(torch.add(output,identity_data)))` I see the normal ResBlock code like this form: `output = self.relu2(torch.add(self.bn2(output),identity_data))` Why Add before batchNorm?

howardchina