SqueezeBERT: What can computer vision teach NLP about efficient neural networks?

TL;DR

SqueezeBERT is a new NLP model inspired by computer vision techniques, specifically grouped convolutions, achieving 4.3 times faster inference on mobile devices while maintaining competitive accuracy.

Contribution

The paper introduces SqueezeBERT, a novel architecture that replaces self-attention operations with grouped convolutions to improve efficiency on mobile devices.

Findings

SqueezeBERT runs 4.3x faster than BERT-base on Pixel 3.

SqueezeBERT achieves competitive accuracy on the GLUE test set.

The approach leverages computer vision techniques for NLP model efficiency.

Abstract

Humans read and write hundreds of billions of messages every day. Further, due to the availability of large datasets, large computing systems, and better neural network models, natural language processing (NLP) technology has made significant strides in understanding, proofreading, and organizing these messages. Thus, there is a significant opportunity to deploy NLP in myriad applications to help web users, social networks, and businesses. In particular, we consider smartphones and other mobile devices as crucial platforms for deploying NLP models at scale. However, today's highly-accurate NLP neural network models such as BERT and RoBERTa are extremely computationally expensive, with BERT-base taking 1.7 seconds to classify a text snippet on a Pixel 3 smartphone. In this work, we observe that methods such as grouped convolutions have yielded significant speedups for computer vision networks, but many of these techniques have not been adopted by NLP neural network designers. We demonstrate how to replace several operations in self-attention layers with grouped convolutions, and we use this technique in a novel network architecture called SqueezeBERT, which runs 4.3x faster than BERT-base on the Pixel 3 while achieving competitive accuracy on the GLUE test set. The SqueezeBERT code will be released.