In diffusion model, denoise network in U-net structure for multi-classes #259
Description
The loss is too small, only 2.xxx then to 1.xx
Is this possible? I checked my code, I thought it was correct. But the loss is too small make me doubt.
{{ message }}