Topology-Aware Collective Algorithm Synthesizer for Distributed Machine Learning
TACOS is an automated synthesizer that generates topology-aware collective algorithms for common distributed machine learning collectives across arbitrary input network topologies.