LoopServe: An Adaptive Dual-phase LLM Inference Acceleration System for Multi-Turn Dialogues
This work addresses efficiency issues for chatbots and virtual assistants by providing an adaptive acceleration method, though it is incremental as it builds on existing context compression and caching techniques.
The paper tackles the computational and memory challenges of large language models in multi-turn dialogues by introducing LoopServe, an adaptive dual-phase inference acceleration system that dynamically selects important attention parts and compresses key value caches, achieving superior effectiveness and significantly faster inference across eleven multi-turn datasets.
Multi-turn dialogues are essential in many real-world applications of large language models, such as chatbots and virtual assistants. As conversation histories become longer, existing large language models face increasing computational and memory challenges, which hinder their ability to provide efficient and responsive interactions. Most current acceleration methods either compress the context or optimize key value caching, but they often rely on fixed or position-based heuristics that do not adapt well to the dynamic and unpredictable patterns found in actual multi-turn conversations. As a result, these models cannot accurately identify and prioritize the most relevant context, leading to degraded response quality. In this paper, we present LoopServe, an adaptive dual-phase inference acceleration framework for large language models in multi-turn dialogues. LoopServe introduces two main innovations. First, it performs online sparsification during the prefilling phase by dynamically selecting the most important parts of the attention matrix for each new input. Second, it uses progressive key value compression during decoding by adaptively maintaining a relevant and efficient cache based on the most recently generated output tokens. We also propose a new benchmark with eleven multi-turn datasets that reflect realistic query positions and conversational dependencies. Extensive experiments demonstrate that LoopServe consistently achieves superior effectiveness compared to existing baselines and significantly accelerates LLM inference across a wide range of long-context dialogue tasks.