task for super-resolution
Opened this issue · 0 comments
poxiao0 commented
Thanks for your work and I have reconduct this model. Now I want to do it for super-resolution task, Can I need to increase the step T? Can I need to replace the adaptive groupnorm with goupnorm because the latent embedding z is for diversity which may not suit for super-resolution task? There are my thoughts but I do not if it is true.