2021/09/15 by Chenhe Dong, Dong, Chenhe, Guangrun Wang +9 · 2 citations
Computer Science · Mathematics · #Artificial intelligence #Artificial neural network #Base (topology) #Benchmark (surveying) #Code (set theory) #Computer science #Distillation #Enhanced Data Rates for GSM Evolution #Inference #Machine learning #Mathematics #Multimodal Machine Learning Applications #Natural Language Processing Techniques #Perceptron #Programming language #Topic Modeling #cs.AI #cs.CL #cs.LG
paper · pdf · doi:10.48550/arxiv.2109.07222
published in arXiv (Cornell University) (Cornell University) · Findings of EMNLP 2021
openalex publication_date 2021/09/15 · arxiv created 2021/09/16 · arxiv updated 2021/09/17 · openalex created_date 2021/11/22 · openalex updated_date 2026/08/08
Pre-trained language models have shown remarkable results on various NLP tasks. Nevertheless, due to their bulky size and slow inference speed, it is hard to deploy them on edge devices. In this paper, we have a critical insight that improving the feed-forward network (FFN) in BERT has a higher gain than improving the multi-head attention (MHA) since the computational cost of FFN is 2∼3 times larger than MHA. Hence, to compact BERT, we are devoted to designing efficient FFN as opposed to previous works that pay attention to MHA. Since FFN comprises a multilayer perceptron (MLP) that is essential in BERT optimization, we further design a thorough search space towards an advanced MLP and perform a coarse-to-fine mechanism to search for an efficient BERT architecture. Moreover, to accelerate searching and enhance model transferability, we employ a novel warm-up knowledge distillation strategy at each search stage. Extensive experiments show our searched EfficientBERT is 6.9× smaller and 4.4× faster than BERT\rmBASE, and has competitive performances on GLUE and SQuAD Benchmarks. Concretely, EfficientBERT attains a 77.7 average score on GLUE test, 0.7 higher than MobileBERT\rmTINY, and achieves an 85.3/74.5 F1 score on SQuAD v1.1/v2.0 dev, 3.2/2.7 higher than TinyBERT4 even without data augmentation. The code is released at https://github.com/cheneydon/efficient-bert.