23959

TopicBERT for Energy Efficient Document Classification

Yatin Chaudhary, Pankaj Gupta, Khushbu Saxena, Vivek Kulkarni, Thomas Runkler, Hinrich Schütze
DRIMCo GmbH, Munich, Germany
arXiv:2010.16407 [cs.CL], (15 Oct 2020)

@misc{chaudhary2020topicbert,

   title={TopicBERT for Energy Efficient Document Classification},

   author={Yatin Chaudhary and Pankaj Gupta and Khushbu Saxena and Vivek Kulkarni and Thomas Runkler and Hinrich Schütze},

   year={2020},

   eprint={2010.16407},

   archivePrefix={arXiv},

   primaryClass={cs.CL}

}

Download Download (PDF)   View View   Source Source   

259

views

Prior research notes that BERT’s computational cost grows quadratically with sequence length thus leading to longer training times, higher GPU memory constraints and carbon emissions. While recent work seeks to address these scalability issues at pre-training, these issues are also prominent in fine-tuning especially for long sequence tasks like document classification. Our work thus focuses on optimizing the computational cost of fine-tuning for document classification. We achieve this by complementary learning of both topic and language models in a unified framework, named TopicBERT. This significantly reduces the number of self-attention operations – a main performance bottleneck. Consequently, our model achieves a 1.4x (~40%) speedup with ~40% reduction in CO2 emission while retaining 99.9% performance over 5 datasets.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2020 hgpu.org

All rights belong to the respective authors

Contact us: