I have a small python app using DiffusionPipeline.from_pretrained
If i run it on localhost the pipeline uses my computers GPU to generate images.
Next i want to host this app in a docker container on HF spaces.
But now the problem is, the pipeline no longer runs on my own GPU but on the HF spaces GPU.
How can i allow my sites visitors to use their own GPU for inference, while also hosting the app containerised? Is this possible? The reason i want to do this is to allow users to use their own GPU instead of me/them paying for cloud hours.