LeanK: Learnable K Cache Channel Pruning for Efficient Decoding

TL;DR

LeanK is a learnable method for pruning key cache channels in large language models, significantly reducing memory usage and increasing decoding speed while maintaining accuracy.

Contribution

It introduces a novel two-stage training process for static channel pruning in key caches, optimizing efficiency without accuracy loss.

Findings

Up to 70% reduction in K cache memory

16%-18% reduction in V cache memory

1.3x speedup in attention computation

Abstract

Large language models (LLMs) enable long-context tasks but face efficiency challenges due to the growing key-value (KV) cache. We propose LeanK, a learning-based method that prunes unimportant key (K) cache channels by leveraging static channel sparsity. With a novel two-stage training process, LeanK learns channel-wise static mask that could satisfy specific sparsity ratio and hardware alignment requirement. LeanK reduces GPU memory and accelerates decoding without sacrificing accuracy. Experiments demonstrate up to 70% K cache and 16%-18% V cache memory reduction. Custom decoding kernel enables 1.3x speedup for attention computation. We also provide insights into model channels and attention heads during long-context inference by analyzing the learned importance distribution. Our code is available at https://aka.ms/LeanK.