rosinality/swapping-autoencoder-pytorch

Where is the parameter initialization?

hdjsjyl opened this issue · 2 comments

hdjsjyl commented

Hi Author,
Thanks for your excellent work. I have confusion where is the parameter initialization? Maybe I missed it, I just want to make sure the initialization. Thank you!

rosinality commented

I have used equalized learning rates, so parameters are initialized from N(0, 1). Then it is scaled during forward pass. https://github.com/rosinality/swapping-autoencoder-pytorch/blob/master/model.py#L32

hdjsjyl commented

Hi Author,
Thanks for your so quick reply. I got it. Thank you so much!