Reproduce “Model-Distributed Inference for Large Language Models at the Edge” 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
Davide Macario, Hulya Seferoglu, Erdem Koyuncu
MDI-LLM distributes a language model’s layers across connected edge devices, passing intermediate results between them so no single device must hold the entire model. It uses recurrent pipeline parallelism to process different text samples at different stages. On Jetson TX2 boards, TinyLlama’s per-device memory fell from 4.57 GB on two nodes to 3.26 GB on three, while aggregate memory rose.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.