StableDiffusionInpaintPipeline does not work with LMSDiscreteScheduler

[shirayu]

Describe the bug

StableDiffusionInpaintPipeline does not work with LMSDiscreteScheduler

Reason

• This seems to be because of the absence of special operations for LMSDiscreteScheduler as performed in StableDiffusionImg2ImgPipeline

diffusers/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py

Lines 206 to 212 in 429dace

	if isinstance(self.scheduler, LMSDiscreteScheduler):
	timesteps = torch.tensor(
	[num_inference_steps - init_timestep] * batch_size, dtype=torch.long, device=self.device
	)
	else:
	timesteps = self.scheduler.timesteps[-init_timestep]
	timesteps = torch.tensor([timesteps] * batch_size, dtype=torch.long, device=self.device)

diffusers/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py

Line 245 in 429dace

timesteps = torch.tensor([timesteps] * batch_size, dtype=torch.long, device=self.device)

Proposal:

I think it is better to unify pipelines as proposed in #551, since most processes are the same.

Reproduction

No response

Logs

"/path/to/lib/python3.10/site-packages/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py", line 245, in __call__
    timesteps = torch.tensor([timesteps] * batch_size, dtype=torch.long, device=self.device)

TypeError: only integer tensors of a single element can be converted to an index

System Info

• diffusers 429dace

[anton-l]

Support for K-LMS is coming in #587