UniPC scheduler does not support new set_timesteps() kwargs

[bghira]

Describe the bug

When using UniPC for SDXL Img2Img pipeline, we run into an error:

    scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
TypeError: set_timesteps() got an unexpected keyword argument 'sigmas'

The function that is called:

def retrieve_timesteps(
    scheduler,
    num_inference_steps: Optional[int] = None,
    device: Optional[Union[str, torch.device]] = None,
    timesteps: Optional[List[int]] = None,
    **kwargs,
):
    """
    Calls the scheduler's `set_timesteps` method and retrieves timesteps from the scheduler after the call. Handles
    custom timesteps. Any kwargs will be supplied to `scheduler.set_timesteps`.

    Args:
        scheduler (`SchedulerMixin`):
            The scheduler to get timesteps from.
        num_inference_steps (`int`):
            The number of diffusion steps used when generating samples with a pre-trained model. If used,
            `timesteps` must be `None`.
        device (`str` or `torch.device`, *optional*):
            The device to which the timesteps should be moved to. If `None`, the timesteps are not moved.
        timesteps (`List[int]`, *optional*):
                Custom timesteps used to support arbitrary spacing between timesteps. If `None`, then the default
                timestep spacing strategy of the scheduler is used. If `timesteps` is passed, `num_inference_steps`
                must be `None`.

    Returns:
        `Tuple[torch.Tensor, int]`: A tuple where the first element is the timestep schedule from the scheduler and the
        second element is the number of inference steps.
    """
    if timesteps is not None:
        accepts_timesteps = "timesteps" in set(
            inspect.signature(scheduler.set_timesteps).parameters.keys()
        )
        if not accepts_timesteps:
            raise ValueError(
                f"The current scheduler class {scheduler.__class__}'s `set_timesteps` does not support custom"
                f" timestep schedules. Please check whether you are using the correct scheduler."
            )
        scheduler.set_timesteps(timesteps=timesteps, device=device, **kwargs)
        timesteps = scheduler.timesteps
        num_inference_steps = len(timesteps)
    else:
        scheduler.set_timesteps(num_inference_steps, device=device, **kwargs)
        timesteps = scheduler.timesteps
    return timesteps, num_inference_steps

seems to check only for timesteps as the missing parameter, but sigmas is required as well

Reproduction

from diffusers import DiffusionPipeline, UniPCMultistepScheduler
import torch

base = DiffusionPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-base-1.0", torch_dtype=torch.float16, variant="fp16", use_safetensors=True
).to("cuda")
base.scheduler = UniPCMultistepScheduler.from_config(base.scheduler.config)

refiner = DiffusionPipeline.from_pretrained(
    "stabilityai/stable-diffusion-xl-refiner-1.0",
    text_encoder_2=base.text_encoder_2,
    vae=base.vae,
    torch_dtype=torch.float16,
    use_safetensors=True,
    variant="fp16",
).to("cuda")
refiner.scheduler = UniPCMultistepScheduler.from_config(refiner.scheduler.config)

prompt = "A majestic lion jumping from a big stone at night"

image = base(
    prompt=prompt,
    num_inference_steps=40,
    denoising_end=0.8,
    output_type="latent",
).images
image = refiner(
    prompt=prompt,
    num_inference_steps=40,
    denoising_start=0.8,
    image=image,
).images[0]

Logs

No response

System Info

Latest git head branch of Diffusers.

Who can help?

No response

[bghira]

well, the lack of check for sigmas was actually because i pulled a version of retrieve_timesteps that lacked that check, somehow - i've updated it so that the check is in my copy as well. i will test some more

[bghira]

adding the check and switching to ddpm have solved the problem for me