Harnessing On-Device Large Language Model: Empirical Results and Implications for AI PC
Harnessing On-Device Large Language Model: Empirical Results and Implications for AI PC
The increasing deployment of Large Language Models (LLMs) on edge devices, driven by model advancements and hardware improvements, offers significant privacy benefits. However, these on-device LLMs inherently face performance limitations due to reduced model capacity and necessary compression techniques. To address this, we introduce a systematic methodology -- encompassing model capability, development efficiency, and system resources -- for evaluating on-device LLMs. Our comprehensive evaluation, encompassing models from 0.5B to 14B parameters and seven post-training quantization (PTQ) methods on commodity laptops, yields several critical insights: 1) System-level metrics exhibit near-linear scaling with effective bits-per-weight (BPW). 2) A practical threshold exists around $\sim$3.5 effective BPW, larger models subjected to low-bit quantization consistently outperform smaller models utilizing higher bit-precision. 3) Quantization with low BPW incurs marginal accuracy loss but significant memory savings. 4) Determined by low-level implementation specifics power consumption on CPU, where computation-intensive operations spend more power than memory-intensive ones. These findings offer crucial insights and practical guidelines for the efficient deployment and optimized configuration of LLMs on resource-constrained edge devices. Our codebase is available at https://github.com/simmonssong/LLMOnDevice.
Qingyu Song、Peiyu Liao、Wenqian Zhao、Yiwen Wang、Shoubo Hu、Hui-Ling Zhen、Ning Jiang、Mingxuan Yuan
计算技术、计算机技术
Qingyu Song,Peiyu Liao,Wenqian Zhao,Yiwen Wang,Shoubo Hu,Hui-Ling Zhen,Ning Jiang,Mingxuan Yuan.Harnessing On-Device Large Language Model: Empirical Results and Implications for AI PC[EB/OL].(2025-05-20)[2025-07-21].https://arxiv.org/abs/2505.15030.点此复制
评论