Subspace Networks: Scaling Decentralized Training with Communication-Efficient Model Parallelism

Open in new window