Try to do inversion,but failed（New in AIGC）
#12
by
FlightRat
- opened
Excuse me,I'm new in AIGC,and I encounter some problems.
I try to modify the pipeline,so that I can do inversion conditioning on the image embedding.But when I turn the result latent back to image,
I got this:
However,if I change the unet of the pipeline by "self.unet = UNet2DConditionModel.from_pretrained("stabilityai/stable-diffusion-2-1-base", subfolder="unet")",then I can get a normal noisy image:
Another question is that did you use anything like adapter during the fine-tune?Or you just fine-tune the whole diffusion model?