BriefGPT.xyz
Dec, 2020
LRC-BERT:用于自然语言理解的潜在表示对比知识蒸馏
LRC-BERT: Latent-representation Contrastive Knowledge Distillation for Natural Language Understanding
HTML
PDF
Hao Fu, Shaojun Zhou, Qihong Yang, Junjie Tang, Guiquan Liu...
TL;DR
本文提出了一种基于对比学习的知识蒸馏方法LRC-BERT,并引入渐变扰动训练架构以提高其鲁棒性。通过验证GLUE基准测试上的8个数据集,表明该方法的性能优于现有最先进的方法,证明了该方法的有效性。
Abstract
The
pre-training models
such as BERT have achieved great results in various
natural language processing
problems. However, a large number of parameters need significant amounts of memory and the consumption of in
→