Fetching the paper…

Maximizing Communication Efficiency for Large-scale Training via 0/1 Adam · Around