UDApter -- Efficient Domain Adaptation Using Adapters

Bhavitvya Malik, Abhinav Ramesh Kashyap, Min-Yen Kan, Soujanya Poria
2/7/2023
cs.CL

Official code: declare-lab/domadapter (10 stars, last commit 2023-02-28).

Abstract

We propose two methods to make unsupervised domain adaptation (UDA) more parameter efficient using adapters, small bottleneck layers interspersed with every layer of the large-scale pre-trained language model (PLM). The first method deconstructs UDA into a two-step process: first by adding a domain adapter to learn domain-invariant information and then by adding a task adapter that uses domain-invariant information to learn task representations in the source domain. The second method jointly learns a supervised classifier while reducing the divergence measure. Compared to strong baselines, our simple methods perform well in natural language inference (MNLI) and the cross-domain sentiment classification task. We even outperform unsupervised domain adaptation methods such as DANN and DSN in sentiment classification, and we are within 0.85% F1 for natural language inference task, by fine-tuning only a fraction of the full model parameters. We release our code at https://github.com/declare-lab/domadapter

View on arXivView PDF

Code Implementations(1)

declare-lab/domadapterOfficial100%

Code for EACL'23 paper "Udapter: Efficient Domain Adaptation Using Adapters"

101Shell, PythonFeb 5, 20233 years ago
adaptersdomain-adaptationparameter-efficient-learningparameter-efficient-tuning

Similar Papers

AdapterSoup: Weight Averaging to Improve Generalization of Pretrained Language Models

87% match2023

Meta-Learning the Difference: Preparing Large Language Models for Efficient Adaptation

85% match1 repo72022

How Useful is Continued Pre-Training for Generative Unsupervised Domain Adaptation?

85% match2024

Mixture-of-Domain-Adapters: Decoupling and Injecting Domain Knowledge to Pre-trained Language Models Memories

85% match1 repo512023

Incremental Pseudo-Labeling for Black-Box Unsupervised Domain Adaptation

85% match2024

Unsupervised Domain Adaptation Using Compact Internal Representations

85% match2024

Cite this paper

@article{malik2023udapter,
  title  = {UDApter -- Efficient Domain Adaptation Using Adapters},
  author = {Bhavitvya Malik and Abhinav Ramesh Kashyap and Min-Yen Kan and Soujanya Poria},
  year   = {2023},
  eprint = {2302.03194},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2302.03194v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2302.03194.svg)](https://codeofpaper.com/paper/2302.03194)

Discussion