Lightweight Adaptation of Neural Language Models via Subspace Embedding

TL;DR

This paper introduces a compact subspace embedding method for neural language models that significantly reduces memory usage with minimal accuracy loss, enabling efficient deployment on resource-constrained devices.

Contribution

The authors propose a novel subspace embedding structure that compresses language model embeddings by over 99.8%, maintaining performance on key NLP tasks.

Findings

Achieves over 99.8% compression rate compared to original embeddings.

Maintains competitive accuracy on XNLI and GLUE benchmarks.

Effective for multilingual and masked language models.

Abstract

Traditional neural word embeddings are usually dependent on a richer diversity of vocabulary. However, the language models recline to cover major vocabularies via the word embedding parameters, in particular, for multilingual language models that generally cover a significant part of their overall learning parameters. In this work, we present a new compact embedding structure to reduce the memory footprint of the pre-trained language models with a sacrifice of up to 4% absolute accuracy. The embeddings vectors reconstruction follows a set of subspace embeddings and an assignment procedure via the contextual relationship among tokens from pre-trained language models. The subspace embedding structure calibrates to masked language models, to evaluate our compact embedding structure on similarity and textual entailment tasks, sentence and paraphrase tasks. Our experimental evaluation shows that the subspace embeddings achieve compression rates beyond 99.8% in comparison with the original embeddings for the language models on XNLI and GLUE benchmark suites.