Output shape should be same as the input shape in conditional image generation

[mamyou96]

Image-Super-Resolution-via-Iterative-Refinement/model/ddpm_modules/diffusion.py

Line 218 in 01d27a7

shape = x.shape

In this line it requires the input shape be same as the output shape in conditional image generation that is not true. Instead of that, the output shape from the config file can be used to create the initial random image.

[LT1st]

You are right, I should have seen this before I spend days on it. Thanks