memory consumption calculation method for deploying LLM using vllm

[Projoke]

Hi, can you provide the memory consumption calculation method for deploying LLM using vllm, as well as the computational approach for multi-machine and multi-GPU?