How to run this on a pytorch or nvidia docker image?

[ohmeow]

I've tried but still keep getting messages about the vLLM editable pip install not being able to determine if cuda is installed. Replacing this line in the requirements.txt

-e git+https://github.com/vllm-project/vllm.git#egg=vllm

... with this gets things up and running ...

vllm==0.2.5

Also I'm using this image: pytorch/pytorch:2.1.1-cuda12.1-cudnn8-runtime