PAIR-Diffusion
PAIR-Diffusion copied to clipboard
hugging face demo can not opened
Runtime error Scheduling failure: not enough hardware capacity
Same!
To fix the hardware capacity issue you need to duplicate the space and assign a GPU instance. I duplicated the space and needed to make the following changes to get it to work again: i) Specify that you need torchaudio==0.10.0+cu113 instead of torchaudio==0.10.0 ii) RUN pip install 'git+https://github.com/facebookresearch/detectron2.git' within the docker file AFTER pip installing requirements.txt (but make sure you remove detectron2 from the requirements.txt). The issue here was that PIL deprecated a type that broke detectron2 0.6. iii) gradio might need to be updated to the latest version (gradio==3.36.1) since there was a KeyError
Worth noting: I did this 22 days ago, so who knows if another dependency broke. Also worth noting, I am not part of the team that developed this, so I might be breaking some dependency by doing what I mentioned.
It also might be worth noting that force restarting the space can sometimes just fix it