MiCRO: Near-Zero Cost Gradient Sparsification for Scaling and Accelerating Distributed DNN Training

Open in new window