lpw_stable_diffusion.py broken when using diffusers==0.9.0

[kovtcharov]

Describe the bug

When initializing a diffusers pipeline, the following error occurs. Possibly related to the following updates recently?

  File ".cache/huggingface/modules/diffusers_modules/git/lpw_stable_diffusion.py", line 755, in __call__
    num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
AttributeError: 'EulerAncestralDiscreteScheduler' object has no attribute 'order'

Reproduction

Simply initializing pipeline with any scheduler (even default scheduler) seems to generate the above error.

Logs

No response

System Info

• diffusers version: 0.9.0
• Platform: Linux-5.15.0-56-generic-x86_64-with-glibc2.35
• Python version: 3.9.15
• PyTorch version (GPU?): 1.13.0+cu117 (True)
• Huggingface_hub version: 0.11.1
• Transformers version: 4.25.1
• Using GPU in script?: yes
• Using distributed or parallel set-up in script?: parallel

[patrickvonplaten]

Gently pinging @SkyTNT here

[kovtcharov]

I tried to reproduce the error in a colab, but ran into a different issue/error related to the SafetyChecker. This was the example lifted from the community page. I'll keep playing around with it:

[~/.cache/huggingface/modules/diffusers_modules/git/lpw_stable_diffusion.py](https://localhost:8080/#) in <module>
      9 from diffusers import SchedulerMixin, StableDiffusionPipeline
     10 from diffusers.models import AutoencoderKL, UNet2DConditionModel
---> 11 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput, StableDiffusionSafetyChecker
     12 from diffusers.utils import PIL_INTERPOLATION, deprecate, logging
     13 from transformers import CLIPFeatureExtractor, CLIPTextModel, CLIPTokenizer

ImportError: cannot import name 'StableDiffusionSafetyChecker' from 'diffusers.pipelines.stable_diffusion' (/usr/local/lib/python3.8/dist-packages/diffusers/pipelines/stable_diffusion/__init__.py)

Looks like it's using the newer diffusers library (0.10.0):

- `diffusers` version: 0.10.0
- Platform: Linux-5.10.133+-x86_64-with-glibc2.27
- Python version: 3.8.16
- PyTorch version (GPU?): 1.13.0+cu116 (False)
- Huggingface_hub version: 0.11.1
- Transformers version: 4.25.1

[kovtcharov]

@patrickvonplaten, would you like me to create a new bug report related to the bug above for diffusers v0.10.0?

[kovtcharov]

Updated script here to reproduce the original bug using diffusers v0.9.0. You should still be able to reproduce the second bug listed above by uncommenting the second line below:

!pip install --upgrade diffusers[torch]==0.9.0 transformers scipy
# !pip install --upgrade diffusers[torch]==0.10.0 transformers scipy

[WASasquatch]

further shenanigans with v0.8.0: #1616****

[patrickvonplaten]

This should be resolved now no?