decentralizepy is a framework for running distributed applications (particularly ML) on top of arbitrary topologies (decentralized, federated, parameter server).
It was primarily conceived for assessing scientific ideas on several aspects of distributed learning (communication efficiency, privacy, data heterogeneity etc.).
