A unified library for building, evaluating, and storing speculative decoding algorithms for LLM inference in vLLM
Open on GitHub