Breaking through Deterministic Barriers: Randomized Pruning Mask Generation and Selection

Jianwei Li, Weizhi Gao, Qi Lei, Dongkuan Xu

Submission Type: Regular Long Paper

Submission Track: Efficient Methods for NLP

Keywords: Pruning, Random, Language Models

TL;DR: The paper introduces a new model pruning method using randomly generated masks and an early evaluation pipeline, achieving top performance on GLUE datasets, particularly at high sparsity levels

Abstract: It is widely acknowledged that large and sparse models have higher accuracy than small and dense models under the same model size constraints. This motivates us to train a large model and then remove its redundant neurons or weights by pruning. Most existing works pruned the networks in a deterministic way, the performance of which solely depends on a single pruning criterion and thus lacks variety. Instead, in this paper, we propose a model pruning strategy that first generates several pruning masks in a designed random way. Subsequently, along with an effective mask-selection rule, the optimal mask is chosen from the pool of mask candidates. To further enhance efficiency, we introduce an early mask evaluation strategy, mitigating the overhead associated with training multiple masks. Our extensive experiments demonstrate that this approach achieves state-of-the-art performance across eight datasets from GLUE, particularly excelling at high levels of sparsity.

Submission Number: 4679