SpeCache: Speculative Key-Value Caching for Efficient Generation of LLMs

Open in new window