Tiny-Attention Adapter: Contexts Are More Important Than the Number of Parameters

TL;DR

This paper introduces tiny-attention adapters for pretrained language models, which use small attention modules to improve transfer learning efficiency and effectiveness, outperforming existing methods with minimal parameter updates.

Contribution

The paper proposes tiny-attention adapters that utilize attention mechanisms with small per-head dimensions, offering a novel approach to parameter-efficient transfer learning.

Findings

Outperforms other parameter-efficient methods on GLUE

Achieves comparable results to GPT-3 and PET on FewGLUE

Uses only 0.05% of parameters for tuning

Abstract

Adapter-tuning is a paradigm that transfers a pretrained language model to downstream tasks by adding and tuning a small number of new parameters. Previously proposed adapter architectures are all feed-forward neural networks. In this paper, we investigate the effectiveness of using tiny-attention -- i.e., attention with extremely small per-head dimensionality -- as adapters. Our tiny-attention adapter learns to modify the hidden states at each position directly conditioned on the hidden states at all the other positions, which is missed by the previously proposed adapters. Moreover, we view its multiple attention heads as a mixture of experts and propose to average their weights during deployment, which further reduces its inference computation cost. On the GLUE benchmark, our tiny-attention adapter outperforms the other parameter-efficient transfer learning methods as well as full fine-tuning while only updating 0.05% of the parameters. On the FewGLUE benchmark, its performance is comparable to that of GPT-3 and PET.