Dist For a node with 4 GPU, run FedAvg with torchrun --nproc_per_node 4 train.py The hyperparameters can be set in train.py