Hello people, I was going through the stable diffusion deep dive notebook,
and am a bit confused as to why noise is being added during “inference”.
I understand that it needs to be added during training but why inference?
Am I missing something here? Is just the sigma being used to scale the noise correction and no noise is being added?