02/02/2021

Learning Light-Weight Translation Models from Deep Transformer

Bei Li, Ziyang Wang, Hui Liu, Quan Du, Tong Xiao, Chunliang Zhang, Jingbo Zhu

Keywords:

Abstract Paper Similar Papers

Abstract: Recently, deep models have shown tremendous improvements in neural machine translation (NMT). However, systems of this kind are computationally expensive and memory intensive. In this paper, we take a natural step towards learning strong but light-weight NMT systems. We proposed a novel group-permutation based knowledge distillation approach to compressing the deep Transformer model into a shallow model. The experimental results on several benchmarks validate the effectiveness of our method. Our compressed model is 8 times shallower than the deep model, with almost no loss in BLEU. To further enhance the teacher model, we present a Skipping Sub-Layer method to randomly omit sub-layers to introduce perturbation into training, which achieves a BLEU score of 30.63 on English-German newstest2014. The code is publicly available at https://github.com/libeineu/GPKD.

The video of this talk cannot be embedded. You can watch it here:
https://slideslive.com/38949243
(Link will open in new window)
 0
 0
 0
 0
  Share    
This is an embedded video. Talk and the respective paper are published at AAAI 2021 virtual conference. If you are one of the authors of the paper and want to manage your upload, see the question "My papertalk has been externally embedded..." in the FAQ section.

Comments

Post Comment
no comments yet
code of conduct: tbd Characters remaining: 140

Similar Papers

 0
 0
 0
 0
 12:13 
 0
 0
 0
 0
 3:23 
 0
 0
 0
 0
 10:42 
 0
 0
 0
 0
 13:28 
 0
 0
 0
 0
 4:59 
 0
 0
 0
 0
 3:29 
 0
 0
 0
 0
 13:02 
 0
 0
 0
 0
 11:36 
 0
 0
 0
 0
 5:02 
 0
 0
 0
 0
 8:20 
 0
 0
 0
 0
 7:05 
 0
 0
 0
 0
 13:42 
 0
 0
 0
 0
 15:59 
 0
 0
 0
 0
 5:21 
 0
 0
 0
 0
 16:56 
 0
 0
 0
 0
 6:15 
 0
 0
 0
 0
 5:06 
 0
 0
 0
 0
 13:17 
 0
 0
 0
 0
 5:09 
 0
 0
 0
 0
 10:28 
 0
 0
 0
 0
 5:11 
 0
 0
 0
 0
 15:00 
 0
 0
 0
 0
 8:51 
 0
 0
 0
 0
 3:23 
 0
 0
 0
 0
 3:20 
 0
 0
 0
 0
 7:33 
 0
 0
 0
 0
 7:58 
 0
 0
 0
 0
 3:54 
 0
 0
 0
 0
 12:16 
 0
 0
 0
 0
 3:52 
 0
 0
 0
 0
 12:00 
 0
 0
 0
 0
 11:16 
 0
 0
 0
 0
 16:12 
 0
 0
 0
 0
 4:15 
 0
 0
 0
 0
 3:05 
 0
 0
 0
 0
 14:52 
 0
 0
 0
 0
 12:43 
 0
 0
 0
 0
 19:14 
 0
 1
 0
 0
 5:08 
 0
 0
 0
 0
 5:01