On Optimizing the Communication of Model Parallelism

Open in new window