Install and Use vLLM for Inference on two Sparks does not work

[eugr]

I suggest you have your script to run git clone/git pull from my repository instead of just copying the Dockerfile to your repo. This way it will stay in sync with all the improvements I make there.