Global and Local Prompts Cooperation via Optimal Transport for Federated Learning [CVPR2024]

The implementation of paper Global and Local Prompts Cooperation via Optimal Transport for Federated Learning (CVPR2024).[paper]

Requirements

• Python 3.8+
• Pytorch 1.10.0+

To install requirements:

pip install -r requirements.txt

Data Preparation

Please follow the instructions at dataset.md to prepare the following datasets: Caltech101, OxfordPets, Flowers102, Food101, DTD.

For CIFAR10 and CIFAR100 datasets, please download and unzip data under DATA/ file catalog. Or simply run experiments with CIFAR10/CIFAR100 dataset, the program will download data automatically.

For DomainNet and office-caltech10 datasets, please follow the instructions of Dataset described here.

How to Run

You can run federated_main.py with some specified arguments.

Training

--root takes as input a path to dataset, like caltech101 or oxford_flowers.

--config-file means which config file to use, such as rn50 or vit_b16.

You can select variables like shots, clients by changing cfg or you can change every arguments you like in fedotp_few_shot.sh.

For example

If you want to use FedOTP to train caltech101 dataset with 2 shots, backbone rn50 and total independent non-iid setting. You can specify that:

MODEL=FedOTP
TRAINER=GLP_OT
DATA=caltech101
SHOTS=2
USEALL=False
IID=False

and run bash scripts/fedotp_few_shot.sh

After the experiments, all the results are finished and save to output/.

Citation

If you find our work useful in your research, please consider citing:

@article{li2024global,
  title={Global and Local Prompts Cooperation via Optimal Transport for Federated Learning},
  author={Li, Hongxia and Huang, Wei and Wang, Jingya and Shi, Ye},
  journal={arXiv preprint arXiv:2403.00041},
  year={2024}
}