DuoGPT: Training-free Dual Sparsity through Activation-aware Pruning in LLMs
Ruokai Yin, Yuhang Li, Donghyun Lee, Priyadarshini Panda
Abstract
Large language models (LLMs) deliver strong performance but are difficult to deploy due to high memory and compute costs. While pruning reduces these demands, most methods ignore activation sparsity observed at runtime. We reinterpret activation sparsity as dynamic structured weight sparsity and propose DuoGPT, a unified framework that constructs dual-sparse (spMspV) workloads by combining unstructured weight pruning with activation sparsity. To preserve accuracy, we extend the Optimal Brain Compression (OBC) framework with activation-aware calibration and introduce output residuals from the dense model as correction terms. We further optimize the solution for efficient GPU execution, enabling scalability to billion-parameter LLMs. Evaluations on LLaMA-2 and LLaMA-3 show that DuoGPT outperforms state-of-the-art structured pruning methods by up to 9.17\% accuracy at an iso-speedup of 1.39$\times$ compared to the baseline dense model. Code is available at GitHub.
BibTeX
@inproceedings{
yin2025duogpt,
title={Duo{GPT}: Training-free Dual Sparsity through Activation-aware Pruning in {LLM}s},
author={Ruokai Yin and Yuhang Li and Donghyun Lee and Priyadarshini Panda},
booktitle={The Thirty-ninth Annual Conference on Neural Information Processing Systems},
year={2025},
url={https://openreview.net/forum?id=PjbpL4brUb}
}