Pre-trained Summarization Distillation

TL;DR

This paper compares three distillation methods for large pre-trained summarization models, finding that the 'shrink and fine-tune' approach often outperforms others on certain datasets, with results demonstrated on Pegasus and BART models.

Contribution

It provides a comprehensive comparison of distillation techniques for summarization models, highlighting the effectiveness of the 'shrink and fine-tune' method across different datasets.

Findings

SFT outperforms knowledge distillation on CNN/DailyMail.

Pseudo-labeling performs better on XSUM.

Code and models are publicly available.

Abstract

Recent state-of-the-art approaches to summarization utilize large pre-trained Transformer models. Distilling these models to smaller student models has become critically important for practical use; however there are many different distillation methods proposed by the NLP literature. Recent work on distilling BERT for classification and regression tasks shows strong performance using direct knowledge distillation. Alternatively, machine translation practitioners distill using pseudo-labeling, where a small model is trained on the translations of a larger model. A third, simpler approach is to 'shrink and fine-tune' (SFT), which avoids any explicit distillation by copying parameters to a smaller student model and then fine-tuning. We compare these three approaches for distillation of Pegasus and BART, the current and former state of the art, pre-trained summarization models, and find that SFT outperforms knowledge distillation and pseudo-labeling on the CNN/DailyMail dataset, but under-performs pseudo-labeling on the more abstractive XSUM dataset. PyTorch Code and checkpoints of different sizes are available through Hugging Face transformers here http://tiny.cc/4iy0tz.