SpikeBERT: A Language Spikformer Learned from BERT with Knowledge Distillation
This work addresses energy efficiency in deep learning for natural language processing, offering a domain-specific improvement for SNNs in language tasks.
The authors tackled the performance gap between spiking neural networks (SNNs) and transformer-based models like BERT for language tasks by developing SpikeBERT, a spiking Transformer trained with a two-stage knowledge distillation method, achieving comparable results to BERT on text classification with significantly less energy consumption.
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.