FedCBC PyTorch Implementation

June 17, 2024 · View on GitHub

This repository contains PyTorch implementation code.

Environment

The system I used and tested in

First, install the packages below:

pytorch==1.12.1
torchvision==0.13.1
matplotlib==3.5.3

For CIFAR-100, please add code below in main.py to get a pre-trained ResNet-18 first.

The code will use 5% training data for each class to train.

cifar100_Data_Spliter().train_feature_extractor()

Once the feature_extractor is trained and saved, please modify Line 27 in main.py to load the feature extractor.

feature_extractor = torch.load('./pretrain_models/resnet-forCIFAR100-30.pth')  # for cifar-100

Then FedCBC is ready to start.

Hyper-parameters can be setted in option.py

All you need to do is to execute code below to train:

python main.py --pseudo_samples 50

In Server.py, Line 153 can controll the global epochs for each task.

Line 52 and Line 139 also need to be modify if you want to adjust the global epoch.

This repository is released under the Apache 2.0 license as found in the LICENSE file.