Host-side TCP setting blocks vLLM backend installation

[krissetto]

I'm running Docker Desktop on Windows and wanted to try vLLM.

Running docker model install-runner --backend=vllm --gpu=cuda results in an error if the user previously already enabled host-side TCP support.

What i see is the following:

This can cause some friction if the user doesn't remember about the setting in Docker Desktop, also because docker ps does not show the inference server container.

Disabling the host-side TCP setting in Desktop and re-running the install-runner command solves the issue and the backend installs successfully