Running the model using "pip install auto-gptq" still results in "CUDA extension not installed"

#3
by mvetter - opened

I was following Marc Sun's instructions to get this running, which say to use "pip install auto-gptq".

However when loading the model using AutoModelForCausalLM.from_pretrained() (in my case in Jupyter), I still get the "CUDA extension not installed" warning. Twice actually.

The model does run and produce output though, so I'm not sure if there's actually an issue here.

Sign up or log in to comment