Every inpainting model 'works' but gives a *.safetensors not found error about sd-webui-inpaint-anything

[Uminosachi]

The display issue 'safetensors not found' has been resolved by setting local_files_only=True in the diffusers' from_pretrained function. Therefore, this issue will be closed.

[Uminosachi]

That log also appears in my environment, and I find it puzzling since the file does exist. However, the Inpaint process itself works without any issues, so it doesn't bother me too much.

[magestyx]

Yes, but no matter which model I use the faces are a bit off, making me wonder if the VAE safetensor would help that out some.

[JonyTianming]

When installing previously, everything that needs to be installed shows that the installation is 100% complete,Downloaded several G things.
But,I have a similar situation,I can't get a picture when I inpaint,It works fine only in controlnet mode：

stabilityai/stable-diffusion-2-inpainting
text_encoder\model.safetensors not found

text_encoder\model.safetensors not found
Error no file named config.json found in directory C:\Users\J.........inpainting\snapshots\6ba40839c3c171123b2b863d16caf023e297abb9\unet.

AND This has happened before：
stabilityai/stable-diffusion-2-inpainting
unet\diffusion_pytorch_model.safetensors not found
We have no connection or you passed local_files_only, so force_download is not an accepted option.

[sean-kang]

I see the inpainting and lama models are downloaded in cache directory. This is an unreliable way to store a persistent data such as checkpoint models since they can be removed anytime. Also, it is inconsistent with where the model files are stored in webui.

To be honest, I think the inpainting feature of this extension is redundant because webui already has inpainting UI and the users are likely to have their own inpainting models. It will be better if the segment anything feature is incorporated into webui's inpainting UI.