LadaBERT: Lightweight Adaptation of BERT through Hybrid Model Compression

TL;DR

LadaBERT introduces a hybrid model compression approach combining pruning, factorization, and distillation to make BERT more efficient for online applications, maintaining high accuracy with significantly reduced training costs.

Contribution

The paper presents LadaBERT, a novel hybrid compression method that reduces training overhead while preserving BERT's performance, addressing limitations of existing distillation techniques.

Findings

Achieves state-of-the-art accuracy on multiple datasets.

Reduces training overhead by an order of magnitude.

Combines multiple compression techniques effectively.

Abstract

BERT is a cutting-edge language representation model pre-trained by a large corpus, which achieves superior performances on various natural language understanding tasks. However, a major blocking issue of applying BERT to online services is that it is memory-intensive and leads to unsatisfactory latency of user requests, raising the necessity of model compression. Existing solutions leverage the knowledge distillation framework to learn a smaller model that imitates the behaviors of BERT. However, the training procedure of knowledge distillation is expensive itself as it requires sufficient training data to imitate the teacher model. In this paper, we address this issue by proposing a hybrid solution named LadaBERT (Lightweight adaptation of BERT through hybrid model compression), which combines the advantages of different model compression methods, including weight pruning, matrix factorization and knowledge distillation. LadaBERT achieves state-of-the-art accuracy on various public datasets while the training overheads can be reduced by an order of magnitude.