CommVQ: Commutative Vector Quantization for KV Cache Compression

TL;DR

CommVQ introduces a novel commutative vector quantization method that significantly compresses KV caches in large language models, enabling longer context processing with minimal accuracy loss and reduced memory footprint.

Contribution

The paper proposes a new commutative vector quantization technique with a lightweight encoder and codebook, optimized for efficient decoding integrated into self-attention, achieving high compression with low overhead.

Findings

Reduces FP16 KV cache size by 87.5% with 2-bit quantization.

Enables 1-bit KV cache quantization with minimal accuracy loss.

Allows LLaMA-3.1 8B to process 128K context length on a single GPU.

Abstract

Large Language Models (LLMs) are increasingly used in applications requiring long context lengths, but the key-value (KV) cache often becomes a memory bottleneck on GPUs as context grows. To address this, we propose Commutative Vector Quantization (CommVQ) to significantly reduce memory usage for long-context LLM inference. We first introduce additive quantization with a lightweight encoder and codebook to compress the KV cache, which can be decoded via simple matrix multiplication. To further reduce computational costs during decoding, we design the codebook to be commutative with Rotary Position Embedding (RoPE) and train it using an Expectation-Maximization (EM) algorithm. This enables efficient integration of decoding into the self-attention mechanism. Our approach achieves high accuracy with additive quantization and low overhead via the RoPE-commutative codebook. Experiments on long-context benchmarks and GSM8K show that our method reduces FP16 KV cache size by 87.5% with 2-bit quantization, while outperforming state-of-the-art KV cache quantization methods. Notably, it enables 1-bit KV cache quantization with minimal accuracy loss, allowing a LLaMA-3.1 8B model to run with a 128K context length on a single RTX 4090 GPU. The source code is available at: https://github.com/UMass-Embodied-AGI/CommVQ.