[bug]: System Error with most Karras schedulers when using DucHaiten-Pony-XL (no-score)

[CubeTheThird]

Is there an existing issue for this problem?

• I have searched the existing issues

Operating system

Linux

GPU vendor

AMD (ROCm)

GPU model

RX 6700 XT

GPU VRAM

12GB

Version number

5.4.2

Browser

Firefox 132.0.2

Python dependencies

{
"accelerate": "1.0.1",
"compel": "2.0.2",
"cuda": null,
"diffusers": "0.31.0",
"numpy": "1.26.4",
"opencv": "4.9.0.80",
"onnx": "1.16.1",
"pillow": "10.4.0",
"python": "3.10.13",
"torch": "2.4.1+rocm6.1",
"torchvision": "0.19.1+rocm6.1",
"transformers": "4.41.1",
"xformers": null
}

What happened

Error logged when attempting to use most Karras schedulers. Errors are like this:

[2024-11-24 18:55:46,409]::[InvokeAI]::ERROR --> Error while invoking session 6f07aea6-40b5-4ed0-bd42-af1133b61b39, invocation 8e57df48-c81e-4e5e-ae74-d3599102a320 (denoise_latents): index 36 is out of bounds for dimension 0 with size 36
[2024-11-24 18:55:46,409]::[InvokeAI]::ERROR --> Traceback (most recent call last):
  File ".venv/lib/python3.10/site-packages/invokeai/app/services/session_processor/session_processor_default.py", line 129, in run_node
    output = invocation.invoke_internal(context=context, services=self._services)
  File ".venv/lib/python3.10/site-packages/invokeai/app/invocations/baseinvocation.py", line 300, in invoke_internal
    output = self.invoke(context)
  File ".venv/lib/python3.10/site-packages/invokeai/app/invocations/denoise_latents.py", line 812, in invoke
    return self._old_invoke(context)
  File ".venv/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
  File ".venv/lib/python3.10/contextlib.py", line 79, in inner
    return func(*args, **kwds)
  File ".venv/lib/python3.10/site-packages/invokeai/app/invocations/denoise_latents.py", line 1073, in _old_invoke
    result_latents = pipeline.latents_from_embeddings(
  File ".venv/lib/python3.10/site-packages/invokeai/backend/stable_diffusion/diffusers_pipeline.py", line 394, in latents_from_embeddings
    step_output = self.step(
  File ".venv/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 116, in decorate_context
    return func(*args, **kwargs)
  File ".venv/lib/python3.10/site-packages/invokeai/backend/stable_diffusion/diffusers_pipeline.py", line 571, in step
    step_output = self.scheduler.step(noise_pred, timestep, latents, **scheduler_step_kwargs)
  File ".venv/lib/python3.10/site-packages/diffusers/schedulers/scheduling_dpmsolver_multistep.py", line 1074, in step
    prev_sample = self.multistep_dpm_solver_third_order_update(self.model_outputs, sample=sample)
  File ".venv/lib/python3.10/site-packages/diffusers/schedulers/scheduling_dpmsolver_multistep.py", line 928, in multistep_dpm_solver_third_order_update
    self.sigmas[self.step_index + 1],
IndexError: index 36 is out of bounds for dimension 0 with size 36

What you expected to happen

The image to generate normally, like their non-Karras counterparts.

How to reproduce the problem

Obtain the DucHaiten-Pony-XL (no-score), alternative download on HuggingFace

Additional context

The schedulers that failed are as follows:

DEIS Karras
DPM++ 2S Karras
DPM++ 2M Karras
DPM++ 2M SDE Karras
DPM++ 3M Karras
Euler Karras
Heun Karras

Additionally, these two, while not failing outright, resulted in black images:

KDPM2 Karras
KDPM2 Ancestral Karras

Discord username

cubethethird