Abstract: We present a comprehensive framework for evaluating distributed AI training performance across single and multi-datacenter environments. Our approach combines 2 existing simulators, ...