search

w86763777 / pytorch-ddpm

Unofficial PyTorch implementation of Denoising Diffusion Probabilistic Models
Do What The F*ck You Want To Public License
477 stars 60 forks source link

Question about loss #7

Closed zy-charon closed 1 year ago

zy-charon commented 1 year ago

Hello, when you calculate the loss, you divide the output of the trainer by 1000. I would like to ask why it is divided by 1000. Shouldn't you calculate the mean of the difference of all pixels?

w86763777 commented 1 year ago

Suppose you mean this line and this line. I follow the simple loss formulation in paper: image which averages the square norm of the differences.

zy-charon commented 1 year ago

Suppose you mean this line and this line. I follow the simple loss formulation in paper: image which averages the square norm of the differences.

Thanks for the reply, you are right, I got it wrong