Model Compression with Multi-Task Knowledge Distillation for Web-scale Question Answering System
arXiv:1904.09636
Abstract
Deep pre-training and fine-tuning models (like BERT, OpenAI GPT) have demonstrated excellent results in question answering areas. However, due to the sheer amount of model parameters, the inference speed of these models is very slow. How to apply these complex models to real business scenarios becomes a challenging but practical problem. Previous works often leverage model compression approaches to resolve this problem. However, these methods usually induce information loss during the model compression procedure, leading to incomparable results between compressed model and the original model. To tackle this challenge, we propose a Multi-task Knowledge Distillation Model (MKDM for short) for web-scale Question Answering system, by distilling knowledge from multiple teacher models to a light-weight student model. In this way, more generalized knowledge can be transferred. The experiment results show that our method can significantly outperform the baseline methods and even achieve comparable results with the original teacher models, along with significant speedup of model inference.
9 pages, 2 figures
References in corpus (5)
- Distilling the Knowledge in a Neural Network
- Exploiting Linear Structure Within Convolutional Networks for Efficient Evaluation
- Speeding up Convolutional Neural Networks with Low Rank Expansions
- Deep Identity-aware Transfer of Facial Attributes
- Semantic Modelling with Long-Short-Term Memory for Information Retrieval
Cited by in corpus (6)
- Well-Read Students Learn Better: On the Importance of Pre-training Compact Models
- Structured Pruning of a BERT-based Question Answering Model
- TwinBERT: Distilling Knowledge to Twin-Structured BERT Models for Efficient Retrieval
- AutoFreeze: Automatically Freezing Model Blocks to Accelerate Fine-tuning
- LightPAFF: A Two-Stage Distillation Framework for Pre-training and Fine-tuning
- On Estimating the Training Cost of Conversational Recommendation Systems