MEMORY-VQ: Compression for Tractable Internet-Scale Memory

TL;DR

MEMORY-VQ introduces a vector quantization approach to compress token representations in memory-augmented language models, significantly reducing storage needs while maintaining performance, thus enabling scalable retrieval augmentation.

Contribution

The paper presents MEMORY-VQ, a novel compression method using VQ-VAE to reduce memory storage in retrieval-augmented models without performance loss.

Findings

Achieves 16x compression rate with LUMEN-VQ.

Maintains comparable performance on KILT benchmark.

Enables practical retrieval augmentation for large corpora.

Abstract

Retrieval augmentation is a powerful but expensive method to make language models more knowledgeable about the world. Memory-based methods like LUMEN pre-compute token representations for retrieved passages to drastically speed up inference. However, memory also leads to much greater storage requirements from storing pre-computed representations. We propose MEMORY-VQ, a new method to reduce storage requirements of memory-augmented models without sacrificing performance. Our method uses a vector quantization variational autoencoder (VQ-VAE) to compress token representations. We apply MEMORY-VQ to the LUMEN model to obtain LUMEN-VQ, a memory model that achieves a 16x compression rate with comparable performance on the KILT benchmark. LUMEN-VQ enables practical retrieval augmentation even for extremely large retrieval corpora.