AttributeError when caching model from command line

[venzen]

I used the following code from this repo's README to cache the stable-diffusion-v1-5-inpainting model from here:

https://huggingface.co/benjamin-paine/stable-diffusion-v1-5-inpainting

from diffusers import StableDiffusionInpaintPipeline
pipe = StableDiffusionInpaintPipeline.from_pretrained("benjamin-paine/stable-diffusion-v1-5-inpainting")
exit()

I had to login with a Huggingface token via command line first and then all files were successfully downloaded. However, while loading pipeline components I got this AttributeError:

  File "/opt/stable-diffusion-webui/venv/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py", line 114, in _inner_fn
    return fn(*args, **kwargs)
  File "/opt/stable-diffusion-webui/venv/lib/python3.10/site-packages/diffusers/pipelines/pipeline_utils.py", line 876, in from_pretrained
    loaded_sub_model = load_sub_model(
  File "/opt/stable-diffusion-webui/venv/lib/python3.10/site-packages/diffusers/pipelines/pipeline_loading_utils.py", line 700, in load_sub_model
    loaded_sub_model = load_method(os.path.join(cached_folder, name), **loading_kwargs)
  File "/opt/stable-diffusion-webui/venv/lib/python3.10/site-packages/transformers/modeling_utils.py", line 2604, in from_pretrained
    state_dict = load_state_dict(resolved_archive_file)
  File "/opt/stable-diffusion-webui/venv/lib/python3.10/site-packages/transformers/modeling_utils.py", line 452, in load_state_dict
    if metadata.get("format") not in ["pt", "tf", "flax"]:
AttributeError: 'NoneType' object has no attribute 'get'

Not sure whether this is caused by a bug in modeling_utils.py or something not-normal in the repo.

[Uminosachi]

Based on the error message, it seems that the model's safetensors file lacks metadata, leading to a None value being returned. This suggests there may be an issue with the file itself.

[venzen]

I found this utility to write metadata to a safetensors file:

https://github.com/by321/safetensors_util

Since it'll be difficult to find accurate metadata for this v1.5 inpainting model, I will inject generic made-up metadata to help modeling_utils.py to read the file.

If I do this, will it affect the functioning of Inpaint Anything ?