An All-Reduce Compatible Top-K Compressor for Communication-Efficient Distributed Learning

Open in new window