geevegeorge / stable-craiyon Goto Github PK

Tried the colab notebook; the Craiyon part works fine, but the Stable Diffusion one throws an error before I could see my prompt Diffusion-ized.

Here is the error:

AttributeError Traceback (most recent call last)

in
1 generator = torch.Generator(device=device).manual_seed(1024)
2 with autocast("cuda"):
----> 3 images = pipe(prompt=prompt, init_image=init_image, strength=0.75, guidance_scale=7.5, generator=generator)["sample"]

1 frames

in call(self, prompt, init_image, strength, num_inference_steps, guidance_scale, eta, generator, output_type)
56
57 # encode the init image into latents and scale the latents
---> 58 init_latents = self.vae.encode(init_image.to(self.device)).sample()
59 init_latents = 0.18215 * init_latents
60

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

I found a similar problem in a different notebook.
There they responded:

We need to update this notebook for diffusers==0.3.0!
For now could you replace:
latents = vae.encode(batch["pixel_values"]).sample().detach()
with
latents = vae.encode(batch["pixel_values"]).latent_dist.sample().detach()

but I couldn't find out what to change in this Stable Craiyon notebook

"Your session crashed for an unknown reason."
WARNING:root:kernel a2d071e6-d71d-4338-8dec-352d2617cda3 restarted