Life of an inference request (vLLM V1): How LLMs are served efficiently at scaleubicloud.com175 pointssamaysharmaa year ago