Enregistré dans:
Détails bibliographiques
Auteurs principaux: Jia, Fucheng, Wu, Zewen, Jiang, Shiqi, Jiang, Huiqiang, Zhang, Qianxi, Yang, Yuqing, Liu, Yunxin, Ren, Ju, Zhang, Deyu, Cao, Ting
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2504.08378
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866911170476113920
author Jia, Fucheng
Wu, Zewen
Jiang, Shiqi
Jiang, Huiqiang
Zhang, Qianxi
Yang, Yuqing
Liu, Yunxin
Ren, Ju
Zhang, Deyu
Cao, Ting
author_facet Jia, Fucheng
Wu, Zewen
Jiang, Shiqi
Jiang, Huiqiang
Zhang, Qianxi
Yang, Yuqing
Liu, Yunxin
Ren, Ju
Zhang, Deyu
Cao, Ting
contents Large language models (LLMs) are increasingly being deployed on mobile devices, but the limited DRAM capacity constrains the deployable model size. This paper introduces ActiveFlow, the first LLM inference framework that can achieve adaptive DRAM usage for modern LLMs (not ReLU-based), enabling the scaling up of deployable model sizes. The framework is based on the novel concept of active weight DRAM-flash swapping and incorporates three novel techniques: (1) Cross-layer active weights preloading. It uses the activations from the current layer to predict the active weights of several subsequent layers, enabling computation and data loading to overlap, as well as facilitating large I/O transfers. (2) Sparsity-aware self-distillation. It adjusts the active weights to align with the dense-model output distribution, compensating for approximations introduced by contextual sparsity. (3) Active weight DRAM-flash swapping pipeline. It orchestrates the DRAM space allocation among the hot weight cache, preloaded active weights, and computation-involved weights based on available memory. Results show ActiveFlow achieves the performance-cost Pareto frontier compared to existing efficiency optimization methods.
format Preprint
id arxiv_https___arxiv_org_abs_2504_08378
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Scaling Up On-Device LLMs via Active-Weight Swapping Between DRAM and Flash
Jia, Fucheng
Wu, Zewen
Jiang, Shiqi
Jiang, Huiqiang
Zhang, Qianxi
Yang, Yuqing
Liu, Yunxin
Ren, Ju
Zhang, Deyu
Cao, Ting
Machine Learning
Large language models (LLMs) are increasingly being deployed on mobile devices, but the limited DRAM capacity constrains the deployable model size. This paper introduces ActiveFlow, the first LLM inference framework that can achieve adaptive DRAM usage for modern LLMs (not ReLU-based), enabling the scaling up of deployable model sizes. The framework is based on the novel concept of active weight DRAM-flash swapping and incorporates three novel techniques: (1) Cross-layer active weights preloading. It uses the activations from the current layer to predict the active weights of several subsequent layers, enabling computation and data loading to overlap, as well as facilitating large I/O transfers. (2) Sparsity-aware self-distillation. It adjusts the active weights to align with the dense-model output distribution, compensating for approximations introduced by contextual sparsity. (3) Active weight DRAM-flash swapping pipeline. It orchestrates the DRAM space allocation among the hot weight cache, preloaded active weights, and computation-involved weights based on available memory. Results show ActiveFlow achieves the performance-cost Pareto frontier compared to existing efficiency optimization methods.
title Scaling Up On-Device LLMs via Active-Weight Swapping Between DRAM and Flash
topic Machine Learning
url https://arxiv.org/abs/2504.08378