ThinK: Thinner Key Cache by Query-Driven Pruning

TL;DR

ThinK introduces a query-dependent pruning method for KV caches in LLMs, significantly reducing memory usage while maintaining accuracy, enabling more efficient long-sequence processing.

Contribution

The paper presents a novel pruning technique that reduces KV cache memory in LLMs by over 20% without sacrificing performance, addressing redundancy in the channel dimension.

Findings

Over 20% reduction in KV cache memory costs.

Achieves 2.8x peak memory reduction with similar accuracy.

Enables up to 5x batch size increase on a single GPU.

Abstract

Large Language Models (LLMs) have revolutionized the field of natural language processing, achieving unprecedented performance across a variety of applications. However, their increased computational and memory demands present significant challenges, especially when handling long sequences. This paper focuses on the long-context scenario, addressing the inefficiencies in KV cache memory consumption during inference. Unlike existing approaches that optimize the memory based on the sequence length, we identify substantial redundancy in the channel dimension of the KV cache, as indicated by an uneven magnitude distribution and a low-rank structure in the attention weights. In response, we propose ThinK, a novel query-dependent KV cache pruning method designed to minimize attention weight loss while selectively pruning the least significant channels. Our approach not only maintains or enhances model accuracy but also achieves a reduction in KV cache memory costs by over 20% compared with vanilla KV cache eviction and quantization methods. For instance, ThinK integrated with KIVI can achieve a 2.8x reduction in peak memory usage while maintaining nearly the same quality, enabling up to a 5x increase in batch size when using a single GPU. Extensive evaluations on the LLaMA and Mistral models across various long-sequence datasets verified the efficiency of ThinK, establishing a new baseline algorithm for efficient LLM deployment without compromising performance. Our code has been made available at https://github.com/SalesforceAIResearch/ThinK.