Reproduce “Dynamic Model Pruning with Feedback” 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
Tao Lin, Sebastian U. Stich, Luis Barba, Daniil Dmitriev, Martin Jaggi
Dynamic Model Pruning with Feedback (DPF) trains a sparse network while retaining a fuller weight state, allowing pruned weights to receive updates and later reactivate. It evaluates gradients on the masked network and periodically changes the mask. On CIFAR-10, WideResNet-28-2 reaches 88.92 ± 0.29% accuracy at 99% target sparsity, exceeding SM and SNIP; this does not establish faster inference.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.