Skip to content

Latest commit

 

History

History
9 lines (6 loc) · 469 Bytes

200205 K-Adapter.md

File metadata and controls

9 lines (6 loc) · 469 Bytes

https://arxiv.org/abs/2002.01808

K-Adapter: Infusing Knowledge into Pre-Trained Models with Adapters (Ruize Wang, Duyu Tang, Nan Duan, Zhongyu Wei, Xuanjing Huang, Jianshu ji, Guihong Cao, Daxin Jiang, Ming Zhou)

요즘 많이 하는 어댑터 붙이기.

  1. LM을 multitask에 프리트레이닝해보고 싶은데 catastrophic forgetting이 신경쓰임.
  2. LM은 고정하고 adapter를 붙여서 각 task에 트레이닝하자.

#language_model #multitask #adapter