Preserving Near-Optimal Gradient Sparsification Cost for Scalable Distributed Deep Learning

Re-assign community