The implementation of federated average learning [1] based on PyTorch.
- GPU RTX3090 + torch1.7.1(cu110) + torchvision 0.8.0
Before run the code, you need to download train dataset and test data of GTSRB from https://benchmark.ini.rub.de/?section=gtsrb&subsection=dataset#Downloads and move them to GTSRB dirctory.
You are supposed to prepare the model in ./checkpoints2/, which is required by load_round in sim_compare1.py. We have given two models for testing. Run server.py first to obtain the required model at the specific round in load_round. Run sim_compare1.py to complete detection of backdoor attack.
Run the code python server.py -nc 100 -cf 0.1 -E 5 -B 10 -mn GTSRB -ncomm 300 -iid 0 -lr 0.01 -vf 1 -g 0 python sim_compare1 -nc 100 -cf 0.1 -E 5 -B 10 -mn GTSRB -ncomm 300 -iid 0 -lr 0.1 -vf 1 -g 0
which means there are 100 clients, we randomly select 10 in each communicating round. The data set are allocated in Non-IID way. The epoch and batch size are set to 5 and 10. The learning rate is 0.1, we validate the codes every 1 rounds during the training, training stops after 300 rounds.
[1] Mcmahan H B , Moore E , Ramage D , et al. Communication-Efficient Learning of Deep Networks from Decentralized Data[J]. 2016.