Task-specific Compression for Multi-task Language Models using Attribution-based Pruning
About
Multi-task language models show outstanding performance for various natural language understanding tasks with only a single model. However, these language models utilize an unnecessarily large number of model parameters, even when used only for a specific task. This paper proposes a novel training-free compression method for multi-task language models using a pruning method. Specifically, we use an attribution method to determine which neurons are essential for performing a specific task. We task-specifically prune unimportant neurons and leave only task-specific parameters. Furthermore, we extend our method to be applicable in low-resource and unsupervised settings. Since our compression method is training-free, it uses few computing resources and does not destroy the pre-trained knowledge of language models. Experimental results on the six widely-used datasets show that our proposed pruning method significantly outperforms baseline pruning methods. In addition, we demonstrate that our method preserves performance even in an unseen domain setting.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Visual Reasoning | Commonsense Reasoning | Jaccard Index (J)6.95 | 30 | |
| Visual Reasoning | Physical Reasoning | J Score8.07 | 30 | |
| Visual Reasoning | Quantitative Reasoning | J Score6.94 | 30 | |
| Visual Grounded Reasoning Analysis | Physical domain dataset | Grounding Score5.6 | 5 | |
| Logit Divergence Analysis | Physical reasoning domain (calibration set) | Mean KL Divergence (nats)10.51 | 5 |