Parameter-efficient Multi-task Fine-tuning for Transformers via Shared Hypernetworks
This work addresses the need for efficient knowledge sharing across tasks in fine-tuning for NLP practitioners, though it is incremental as it builds on existing adapter methods.
The paper tackles the problem of parameter-efficient multi-task fine-tuning for transformers by introducing shared hypernetworks to generate adapter parameters across tasks, achieving improved performance on the GLUE benchmark with only 0.29% additional parameters per task and showing substantial gains in few-shot domain generalization.
State-of-the-art parameter-efficient fine-tuning methods rely on introducing adapter modules between the layers of a pretrained language model. However, such modules are trained separately for each task and thus do not enable sharing information across tasks. In this paper, we show that we can learn adapter parameters for all layers and tasks by generating them using shared hypernetworks, which condition on task, adapter position, and layer id in a transformer model. This parameter-efficient multi-task learning framework allows us to achieve the best of both worlds by sharing knowledge across tasks via hypernetworks while enabling the model to adapt to each individual task through task-specific adapters. Experiments on the well-known GLUE benchmark show improved performance in multi-task learning while adding only 0.29% parameters per task. We additionally demonstrate substantial performance improvements in few-shot domain generalization across a variety of tasks. Our code is publicly available in https://github.com/rabeehk/hyperformer.