Stable diffusion pipeline #168
Conversation
|
The documentation is not available anymore as the PR was closed or merged. |
|
Looks good to me! One minor question: since this is only for |
Good remark - think this makes sense! It'll make the code cleaner and we could always add it later to the party |
|
Good point @pcuenca ! I agree with this, but since the original codebase allows to play with both DDIM and PNDM, think it would important to have the |
|
@patrickvonplaten @anton-l
|
src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py
Outdated
Show resolved
Hide resolved
| # expand the latents if we are doing classifier free guidance | ||
| if do_classifier_free_guidance: | ||
| latents = torch.cat((latents, latents), dim=0) |
There was a problem hiding this comment.
since text_embeddings are already expanded, we only need to expand latents here.
There was a problem hiding this comment.
that gave a problem with naming but should be fixed now
src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py
Outdated
Show resolved
Hide resolved
src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py
Outdated
Show resolved
Hide resolved
|
|
||
| for t in tqdm(self.scheduler.timesteps): | ||
| # expand the latents if we are doing classifier free guidance | ||
| latent_model_input = torch.cat([latents] * 2) if do_classifier_free_guidance else latents |
There was a problem hiding this comment.
@patil-suraj there was a bug with the naming here -> corrected it
| if torch_device is None: | ||
| torch_device = "cuda" if torch.cuda.is_available() else "cpu" | ||
|
|
||
| if isinstance(prompt, str): |
There was a problem hiding this comment.
@patil-suraj -> made sure a string can be passed as input
| self.text_encoder.to(torch_device) | ||
|
|
||
| # get prompt text embeddings | ||
| text_input = self.tokenizer(prompt, padding=True, truncation=True, return_tensors="pt") |
There was a problem hiding this comment.
@patil-suraj removed the weird 77 max length here
patrickvonplaten
changed the title
* add stable diffusion pipeline * get rid of multiple if/else * batch_size is unused * add type hints * Update src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py * fix some bugs Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
This PR adds
StableDiffusionPipeline