INSTRUCTION

This code is the implement of the Grad-based extraction sub-model Federated Learning.

DEMO

python train_classifier.py --rounds 100 --lr 0.01 --mode roll --model_name conv --dataset emnist --shardperuser 2 --group_name xxx --client_send_label --device 0

resnet34 + tinyimagenet

awareGrad

python train_classifier.py --rounds 100 --lr 0.01 --mode awareGrad --model_name resnet34 --dataset tinyimagenet --shardperuser 20 --group_name xxx --client_send_label --device 0

roll

python train_classifier.py --rounds 100 --lr 0.01 --mode roll --model_name resnet34 --dataset tinyimagenet --shardperuser 20 --group_name xxx --device 0

parameters mode

awareGrad -> FedGSE

aware -> FedDSE

roll -> FedRolex

rand -> Federated Dropout

hetero -> HeteroFL

fedavg -> FedAVG