Reproduce “Galaxy: A Resource-Efficient Collaborative Edge AI System for In-situ Transformer Inference” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Shengyuan Ye, Jiangsu Du, Liekang Zeng, Wenzhong Ou, Xiaowen Chu, Yutong Lu, Xu Chen
Galaxy enables single-request Transformer inference across nearby edge devices by splitting attention and MLP operations with tensor parallelism and connective operations with sequence parallelism. It profiles devices, plans work allocation, and overlaps communication with computation. On Jetson Nano testbeds, it reports 1.3×–2.5× lower inference latency than parallel baselines in heterogeneous settings at 125 Mbps; results do not establish security benefits.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.