Low Precision Local Training is Enough for Federated Learning | Read Paper on Bytez