Thanks for that. I've used RunPod GPU cloud to setup vLLM as an Open-AI API compatible endpoint before, but haven't tried any of the serverless options yet.