Scaffold-Federated-Learning

PyTorch implementation of SCAFFOLD (Stochastic Controlled Averaging for Federated Learning, ICML 2020).

Environment

numpy==1.18.5

pytorch==1.10.1+cu111

Experimental parameter settings

communication rounds: r=10,

number of local update steps: E=10,

=0.01,

=1,

total number of clients: K=10,

sampled num: |S|=5.

Usage

python main.py