The source code for "Text-Enhanced Data-free Approach for Federated Class-Incremental Learning" accepted by CVPR 2024.

Paper link: https://arxiv.org/abs/2403.14101

Reproducing

We test the code on RTX 4090 GPU with pytorch:

torch==2.0.1
torchvision==0.15.2

Baseline

Here, we provide a simple example for different methods. For example, for cifar100-5tasks, please run the following commands to test the model performance with non-IID ($\beta=0.5$) data.

#!/bin/bash
# method= ["finetue", "lwf", "ewc", "icarl", "target"]

CUDA_VISIBLE_DEVICES=0 python main.py --group=c100t5 --exp_name=$method_b05 --dataset cifar100 --method=$method --tasks=5 --num_users 5 --beta=0.5

Ours

CUDA_VISIBLE_DEVICES=0 python main.py --group=c100t5 --exp_name=lander_b05 --dataset cifar100 --method=lander --tasks=5 --num_users 5 --beta=0.5

Citation:

@article{lander,
title={Text-Enhanced Data-free Approach for Federated Class-Incremental Learning},
author={Tran, Minh-Tuan and Le, Trung and Le, Xuan-May and Harandi, Mehrtash and Phung, Dinh},
journal={arXiv preprint arXiv:2403.14101},
year={2024}
}