We focus on the problem of efficiently deploying a federated learning training task in a decentralized setting with multiple aggregators. To that end, we introduce a number of improvements and modifications to the recently ...