ChrisTiger93

while the validation output during training seems to be good. Any bugs in the inference code ? Or it is due to different diffuser version?