Update lisa_on_cuda/utils/app_helpers.py

#3

The inference_decorator was not passed to the model loading function. This gives the illusion that the model load function has no GPU. It still worked before.
There are other ways to do .to("cuda") the models in the inference function after they have been loaded into the CPU. Where this is possible, this is lighter.

aletrn changed pull request status to merged

Sign up or log in to comment