Progressive Growing in StyleGAN error with size off test_x = test_stylegan_block.adain(test_x, test_w)

Course Q&A Generative Adversarial Networks (GANS) Build Better Generative Adversarial Networks
1

Deleted
Deleted
Deleted

2

Hi, @Stefan_Persson!

in your implementation of MicroStyleGANGeneratorBlock class instead of
if self.use_upsample:
self.upsample = nn.Upsample(size=(starting_size,starting_size), mode=‘bilinear’)
try
if self.use_upsample:
self.upsample = nn.Upsample(starting_size, mode=‘bilinear’)

This is the first significant error that I found. There may be more. Please try this first and share the outcome.

PS. For your future reference, sharing the code implementation in a public thread isn’t recommended.

Kind regards,
DK.

3

Thank you Dmitriy_KhvanGANS Mentor

I have made your change. The error is still the same.

4

Good. In InjectNoise class instead of torch.randn(1,channels,1,1) try torch.randn((1,channels,1,1)).

5

Thank you Dmitriy_KhvanGANS Mentor
I have changed to torch.randn((1,channels,1,1)) in InjectNoise class and it is still the same.

6

Hi Stefan!

I request you not to upload your code ( Kindly delete it ). If you are having serious trouble completing your assignment, share your notebook with any of the mentors through private chat.

7

@Stefan_Persson can you send me your notebook via private chat?

8

Dmitriy_KhvanGANS Mentor

Where is the chat function? I posted it here, since I did not find any.

9

Please click on my tag and a pop-up with Message button will appear.
Screenshot 2023-09-06 at 14.26.02

10

The notebook well received, troubleshooting via private chat.

11

Resolved.