-
If a backward function returns None/ATen undefined tensor, it can be passed to a downstream ATen autograd function as a grad, but ATen autograd functions in general do not check that their grad argume…
-
-
I think our current model with `TensorGraph` can't nicely support GANs. In order to train GANs, you need to train both a discriminator `D` and a generator `G`. The training of these two models is thre…
-
Hi, do you plan to provide a pytorch implementation of [the recent paper](https://arxiv.org/abs/1704.00028) on "Improved Training of Wasserstein GANs"?
Is there an easy way to compute the gradient w.…
-
I’ve met a strange memory leak when I tried to implement “Improved Training of Wasserstein GANs”. I’m getting OOM in the middle of second epoch both on CPU and GPU. Memory usage seems to increase afte…
Randl updated
6 years ago
-
Hi,
I had to fix a couple of things in the code , I presume keras latest version changed some variable definitions, but I haven't tried with older keras versions so I'm not sure about that.
Anyway, …
-
Thanks for sharing the code. The code is elegant and well structured, I'm going to pick this one as a starting point.
But there might be a glitch in the batch normalization.
Are you using batch nor…
-
-
@bobchennan Hi Chennan, glad to see your implementations of WGANs. Thank you for your contribution. When reading your codes, I see two main changes that you have made:
(1) Line 55: K.mean(target\*out…