FastBERT: a Self-distilling BERT with Adaptive Inference Time

Abstract

Pre-trained language models like BERT have proven to be highly performant. However, they are often computationally expensive in many practical scenarios, for such heavy models can hardly be readily implemented with limited resources. To improve their efficiency with an assured model performance, we propose a novel speed-tunable FastBERT with adaptive inference time. The speed at inference can be flexibly adjusted under varying demands, while redundant calculation of samples is avoided. Moreover, this model adopts a unique self-distillation mechanism at fine-tuning, further enabling a greater computational efficacy with minimal loss in performance. Our model achieves promising results in twelve English and Chinese datasets. It is able to speed up by a wide range from 1 to 12 times than BERT if given different speedup thresholds to make a speed-performance tradeoff.

This manuscript has been accepted to appear at ACL 2020

References in corpus (4)

• Distilling the Knowledge in a Neural Network
• FitNets: Hints for Thin Deep Nets
• Compressing Deep Convolutional Networks using Vector Quantization
• ERNIE: Enhanced Representation through Knowledge Integration

Cited by in corpus (5)

• AutoFreeze: Automatically Freezing Model Blocks to Accelerate Fine-tuning
• Co-BERT: A Context-Aware BERT Retrieval Model Incorporating Local and Query-specific Context
• IOT: Instance-wise Layer Reordering for Transformer Structures
• Improving NER's Performance with Massive financial corpus
• Retraining DistilBERT for a Voice Shopping Assistant by Using Universal Dependencies