Continual Sequence Generation with Adaptive Compositional Modules

TL;DR

This paper introduces an adaptive modular approach for continual sequence generation that dynamically adds or reuses transformer modules based on task similarity, improving performance and efficiency.

Contribution

It proposes a novel adaptive compositional module framework with pseudo experience replay for continual sequence generation, balancing knowledge sharing and task-specific adaptation.

Findings

Outperforms state-of-the-art baselines in various sequence generation tasks.

Effectively balances module reuse and addition based on task similarity.

Enhances parameter efficiency while maintaining high performance.

Abstract

Continual learning is essential for real-world deployment when there is a need to quickly adapt the model to new tasks without forgetting knowledge of old tasks. Existing work on continual sequence generation either always reuses existing parameters to learn new tasks, which is vulnerable to catastrophic forgetting on dissimilar tasks, or blindly adds new parameters for every new task, which could prevent knowledge sharing between similar tasks. To get the best of both worlds, in this work, we propose continual sequence generation with adaptive compositional modules to adaptively add modules in transformer architectures and compose both old and new modules for new tasks. We also incorporate pseudo experience replay to facilitate knowledge transfer in those shared modules. Experiment results on various sequences of generation tasks show that our framework can adaptively add modules or reuse modules based on task similarity, outperforming state-of-the-art baselines in terms of both performance and parameter efficiency. We make our code public at https://github.com/GT-SALT/Adaptive-Compositional-Modules.