'is_inpainting_task': 'inpainting' in model.name_or_path

[crapthings]

Traceback (most recent call last):
File "/home/zznet/workspace/ai-pipe/prod-ai-pipe-replace-model/runpod_app.py", line 14, in
from pipe import text2img, getText2imgPipe, set_sampler, compel
File "/home/zznet/workspace/ai-pipe/prod-ai-pipe-replace-model/pipe.py", line 34, in
apply_hidiffusion(text2imgPipe)
File "/home/zznet/.local/lib/python3.10/site-packages/hidiffusion/hidiffusion.py", line 1959, in apply_hidiffusion
'is_inpainting_task': 'inpainting' in model.name_or_path,
TypeError: argument of type 'NoneType' is not iterable

[ShenZhang-Shin]

Do you run inpainting following the code in readme? Or you use your own code

[minipuding]

I have the same problem

[ShenZhang-Shin]

You can check whether model is NoneType. Or can you show your code?

[iwr-redmond]

Same error here.

Code snippets:

base_model_id='/var/storage/checkpoints/softrealistic_v187.safetensors'

self.pipe = StableDiffusionPipeline.from_single_file(
                        base_model_id,
                        torch_dtype=self.type_model_precision,
                    )
                    class_name = "StableDiffusionPipeline"

apply_hidiffusion(self.pipe)

diffusers==0.29.0

[JohnnyStreet]

A solution was already posted here but never added to the repo #3

EDIT: It looks like the codebase has changed where this solution was applied before. I'll try to update and submit a PR today.

from diffusers.pipelines import auto_pipeline
...
def apply_hidiffusion(
        model: torch.nn.Module,
        apply_raunet: bool = True,
        apply_window_attn: bool = True,        
        is_playground = False):
        ...
        'is_inpainting_task': model.__class__ in auto_pipeline.AUTO_INPAINT_PIPELINES_MAPPING.values(), 
        'is_playground': is_playground,

Additionally, it seems to be expecting the pipeline to have a _num_timesteps attribute. Since there was previously a hard-coded value of 50, I assume this is num_inference_steps. I was able to solve this by doing the following before inference:

setattr(image_pipeline, "_num_timesteps", request.num_inference_steps)

After implementing these changes it is working.