PyTorch implementation of SCAFFOLD (Stochastic Controlled Averaging for Federated Learning, ICML 2020).
numpy==1.18.5
pytorch==1.10.1+cu111
communication rounds: r=10,
number of local update steps: E=10,
=0.01,
=1,
total number of clients: K=10,
sampled num: |S|=5.
python main.py