Reproduce “EnerInfer: Energy-Aware On-Device LLM 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
Bohua Zou, Nian Liu, Binqi Sun, Matteo Mascherin, Debayan Roy, Yutao Liu, Yu Peng, Ning Jia and 1 more
EnerInfer predicts model-specific throughput and power across NPU and DDR clock settings, then selects the most energy-efficient setting that meets a decoding-speed target and adjusts control when overheating is forecast. In a high-end phone running LLaMA2-1.3B Q4, it improves decoding tokens per joule by 65% over maximum-frequency default at a 5-token/s target. Real-world tests report 4.2–11% lower total-task energy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.