Non-linear functions are prevalent in Transformers and their lightweight variants, incurring substantial and frequently underestimated hardware costs. Previous state-of-the-art works optimize these operations by piece-wise linear approximation and store the parameters in look-up tables (LUT), but most of them require unfriendly high-precision arithmetics such as FP/INT 32 and lack consideration of integer-only INT quantization. This paper proposed a genetic LUT-Approximation algorithm namely GQA-LUT that can automatically determine the parameters with quantization awareness. The results demonstrate that GQA-LUT achieves negligible degradation on the challenging semantic segmentation task for both vanilla and linear Transformer models. Besides, proposed GQA-LUT enables the employment of INT8-based LUT-Approximation that achieves an area savings of 81.3~81.7% and a power reduction of 79.3~80.2% compared to the high-precision FP/INT 32 alternatives. Code is available at https:// github.com/PingchengDong/GQA-LUT.
@article{arxiv.2403.19591,
title = {Genetic Quantization-Aware Approximation for Non-Linear Operations in Transformers},
author = {Pingcheng Dong and Yonghao Tan and Dong Zhang and Tianwei Ni and Xuejiao Liu and Yu Liu and Peng Luo and Luhong Liang and Shih-Yang Liu and Xijie Huang and Huaiyu Zhu and Yun Pan and Fengwei An and Kwang-Ting Cheng},
journal= {arXiv preprint arXiv:2403.19591},
year = {2024}
}