Reproduce “DynSL: Resource-Adaptive Model Training across Heterogeneous Edge Devices” 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
Boris Radovič, Istemi Ekin Akkus, Marco Canini, Ruichuan Chen, Alice Dethise, Ivica Rimac, Veljko Pejović
DynSL adapts layer placement in federated split learning using client timing, network throughput, and server load to reallocate work. It starts without profiling and moves layers when predicted epoch time improves. When an Orange Pi link was throttled to 10 Mbps five seconds into a ResNet18/CINIC10 epoch, time was 322.4 seconds, versus 981.3 with static DynSL-P. This combines online adaptation with a profiling-free start.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.