SpikeBERT: A Language Spikformer Learned from BERT with Knowledge Distillation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Lv, Changze, Li, Tianlong, Xu, Jianhan, Gu, Chenxi, Ling, Zixuan, Zhang, Cenyuan, Zheng, Xiaoqing, Huang, Xuanjing
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911780950769664
author Lv, Changze
Li, Tianlong
Xu, Jianhan
Gu, Chenxi
Ling, Zixuan
Zhang, Cenyuan
Zheng, Xiaoqing
Huang, Xuanjing
author_facet Lv, Changze
Li, Tianlong
Xu, Jianhan
Gu, Chenxi
Ling, Zixuan
Zhang, Cenyuan
Zheng, Xiaoqing
Huang, Xuanjing
contents Spiking neural networks (SNNs) offer a promising avenue to implement deep neural networks in a more energy-efficient way. However, the network architectures of existing SNNs for language tasks are still simplistic and relatively shallow, and deep architectures have not been fully explored, resulting in a significant performance gap compared to mainstream transformer-based networks such as BERT. To this end, we improve a recently-proposed spiking Transformer (i.e., Spikformer) to make it possible to process language tasks and propose a two-stage knowledge distillation method for training it, which combines pre-training by distilling knowledge from BERT with a large collection of unlabelled texts and fine-tuning with task-specific instances via knowledge distillation again from the BERT fine-tuned on the same training examples. Through extensive experimentation, we show that the models trained with our method, named SpikeBERT, outperform state-of-the-art SNNs and even achieve comparable results to BERTs on text classification tasks for both English and Chinese with much less energy consumption. Our code is available at https://github.com/Lvchangze/SpikeBERT.
format Preprint
id arxiv_https___arxiv_org_abs_2308_15122
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle SpikeBERT: A Language Spikformer Learned from BERT with Knowledge Distillation
Lv, Changze
Li, Tianlong
Xu, Jianhan
Gu, Chenxi
Ling, Zixuan
Zhang, Cenyuan
Zheng, Xiaoqing
Huang, Xuanjing
Computation and Language
Spiking neural networks (SNNs) offer a promising avenue to implement deep neural networks in a more energy-efficient way. However, the network architectures of existing SNNs for language tasks are still simplistic and relatively shallow, and deep architectures have not been fully explored, resulting in a significant performance gap compared to mainstream transformer-based networks such as BERT. To this end, we improve a recently-proposed spiking Transformer (i.e., Spikformer) to make it possible to process language tasks and propose a two-stage knowledge distillation method for training it, which combines pre-training by distilling knowledge from BERT with a large collection of unlabelled texts and fine-tuning with task-specific instances via knowledge distillation again from the BERT fine-tuned on the same training examples. Through extensive experimentation, we show that the models trained with our method, named SpikeBERT, outperform state-of-the-art SNNs and even achieve comparable results to BERTs on text classification tasks for both English and Chinese with much less energy consumption. Our code is available at https://github.com/Lvchangze/SpikeBERT.
title SpikeBERT: A Language Spikformer Learned from BERT with Knowledge Distillation
topic Computation and Language
url https://arxiv.org/abs/2308.15122