Comments (3)
Hi,
In the Residual block, we used BatchNorm (basically, Generator uses it but Discriminator uses Layer Norm).
https://github.com/jalola/improved-wgan-pytorch/blob/master/models/wgan.py#L94
In the Tensorflow implementation, the author also used BatchNorm:
https://github.com/igul222/improved_wgan_training/blob/master/gan_64x64.py#L189
from improved-wgan-pytorch.
Thanks for the links.
And i am confused about
improved-wgan-pytorch/models/conwgan.py
Line 222 in 5bf75f8
whether can we add batch norm after "output_gan", since this branch has no relation with the wasserstein distance.
For example
# in the self.__init__
self.bn = nn.BatchNorm1d(feature_nums)
...
# in the self.forward
output_congan = self.ln2(output)
output_congan = self.bn(output_congan)
from improved-wgan-pytorch.
Sorry for misunderstanding your question. I got it now.
However, I am not experienced in using BatchNorm and I rarely see people use BatchNorm at the end of the output. There, we want the output to be a label (0, 1, 2 ...) of image input so we can compute the loss with real label (https://github.com/jalola/improved-wgan-pytorch/blob/master/congan_train.py#L253).
So I think we don't need to use BatchNorm.
You can try it out and let us know about that.
from improved-wgan-pytorch.
Related Issues (20)
- Why comment aux_errD_fake loss (aux error of fake data during training D)? HOT 3
- RuntimeError: Mismatch in shape: grad_output[0] has a shape of torch.Size([1]) and output[0] has a shape of torch.Size([]) HOT 3
- why doesnt Wasserstein distance in Wgan converge? HOT 1
- Compare to the tensorflow version, this code cost too much time. HOT 2
- Generator loss function HOT 1
- Model Diverge! HOT 2
- cant work :RuntimeError: shape '[32, 3, 96, 96]' is invalid for input of size 580608 HOT 4
- Questions about noise HOT 1
- Backward HOT 4
- One and Mone HOT 4
- Regarding conditioning in the generator HOT 1
- lsun data issue HOT 1
- Increasing dimensionality HOT 3
- Is Wasserstein-distance calculated correctly? HOT 2
- about the size of images HOT 1
- Conditional GAN: NUM_CLASS param for image completion task and whats the deal with random noise? HOT 5
- Some question in model.py HOT 2
- Missing components HOT 1
- A question about loss_G.backward(Mone) HOT 2
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from improved-wgan-pytorch.