Reproduce “Energy-efficient Task Adaptation for NLP Edge Inference Leveraging Heterogeneous Memory Architectures” 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
Zirui Fu, Aleksandre Avaliani, Marco Donato
The authors adapt ALBERT for multi-task inference using small task-specific adapters and a memory design that separates shared from changeable weights. In a modeled three-task comparison at Q3,5, the system reports 5.91× lower area, 682.27× lower energy per inference, and 62.37× lower latency than vanilla ALBERT; QNLI and SST-2 accuracy losses qualify this modeled, not device-measured, finding.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.