Installation

Requirements

• Python 3.6.8+
• Numpy
• Pytorch 1.4.0+
• torchvision
• hydra-core=0.9.0
• tensorboard
• omegaconf=1.3.0

Train

# example
$ python train.py n_round=100 C=0.3

Usages config\config.yaml

• model: vgg

• fed: fedCDP # algorithm

• optim: sgd

• device: 'cuda'

• n_round: 200 # Number of training rounds

• E: 5 # Local epoch

• K: 100 # Number of clients

• C: 0.1 # Percentage of clients participating in training in each round

• B: 64 # Batchsize

• iid: [True, False] # Whether the data is IID distribution

• dataSet: [mnist, cifar10, cifar100]