arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2110.13640cs.CL
收藏

s2s-ft:用于序列到序列学习的微调预训练Transformer编码器

s2s-ft: Fine-Tuning Pretrained Transformer Encoders for Sequence-to-Sequence Learning

Microsoft Research(微软研究院)

Hangbo Bao, Li Dong, Wenhui Wang, Nan Yang, Furu Wei

摘要:

预训练的双向Transformers,例如BERT,在各种各样的语言理解任务中已经实现了显著的改进,而直接将它们应用于自然语言生成并不简单。在本文中,我们提出了一个序列到序列的微调工具包s2 s-ft,它采用预训练的Transformers的条件生成任务。受UniLM的启发,我们实现了三种序列到序列的微调算法,即因果微调、掩蔽微调和伪掩蔽微调。通过利用现有的预训练的双向Transformers,实验结果表明,s2 s-ft在抽象摘要和问题生成的几个基准上实现了强大的性能。此外,我们证明了包s2 s-ft支持单语和多语言NLG任务。s2 s-ft工具包可以在https://github.com/microsoft/unilm/tree/master/s2s-ft上找到。

英文摘要:

Pretrained bidirectional Transformers, such as BERT, have achieved significant improvements in a wide variety of language understanding tasks, while it is not straightforward to directly apply them for natural language generation. In this paper, we present a sequence-to-sequence fine-tuning toolkit s2s-ft, which adopts pretrained Transformers for conditional generation tasks. Inspired by UniLM, we implement three sequence-to-sequence fine-tuning algorithms, namely, causal fine-tuning, masked fine-tuning, and pseudo-masked fine-tuning. By leveraging the existing pretrained bidirectional Transformers, experimental results show that s2s-ft achieves strong performance on several benchmarks of abstractive summarization, and question generation. Moreover, we demonstrate that the package s2s-ft supports both monolingual and multilingual NLG tasks. The s2s-ft toolkit is available at https://github.com/microsoft/unilm/tree/master/s2s-ft.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑