AttributeError: 'AutoencoderKLOutput' object has no attribute 'sample'

#53
by ArchaeonSeq - opened

Started getting this error suddenly even though I have been using the code with no issues for some time

CompVis org

Hi @ArchaeonSeq , there were breaking changes in the scheduler outputs of diffusers version 0.3.0: https://github.com/huggingface/diffusers/releases/tag/v0.3.0.

In this case, I think you can use the following instead:

init_latents = self.vae.encode(init_image.to(self.device)).latent_dist.sample()

or:

init_latents = self.vae.encode(init_image.to(self.device))[0].sample()

Sign up or log in to comment