A Method for Building Large Language Models with Predefined KV Cache Capacity

TL;DR

This paper presents the Bounded-Cache Transformer (BCT), a new method for large language models that efficiently manages KV cache capacity to reduce memory usage without sacrificing performance.

Contribution

The BCT introduces a bounded-length KV cache mechanism for Transformer models, enabling efficient inference with limited memory resources.

Findings

Reduces memory consumption during inference

Maintains inference quality with limited cache capacity

Demonstrates significant efficiency improvements in experiments

Abstract

This paper introduces a novel approach, the Bounded-Cache Transformer (BCT), for building large language models with a predefined Key-Value (KV) cache capacity. The BCT addresses the excessive memory consumption issue in traditional KV caches by implementing a bounded-length KV cache, which is particularly suitable for the attention layers in Transformer decode-only architectures. By dynamically updating the key-value vector sequences, the BCT achieves efficient inference within limited cache capacity, significantly reducing memory usage while maintaining model performance and system throughput. Experimental results demonstrate that the BCT significantly reduces memory usage while maintaining the model's inference quality, offering a new solution for efficient inference in large language models.