FedAPT

This repository is the implementation of "Federated Adaptive Prompt Tuning for Multi-domain Collaborative Learning" in AAAI2024.

The original code has not been organized yet, and we plan to refactor the code in the future to improve readability.

Requirements

Dependencies

Python 3.7.11
torch 1.10.2
torchvision 0.12.0+cu113
tqdm 4.63.0
numpy
clip 1.0

Datasets

A datapath should be defined (such as datapath='/home/share/DomainNet/'). The directory structure should be

/home/share/DomainNet/
│       
└───clipart
│   │...
└───infograph
│   │...
...
└───sketch
│   │...   

Download and unzip the DomainNet dataset to datapath.

Training

Each domain has one client:

PromptFL: python prompt-promptfl.py --logname 'xxxx' --datapath 'xxxx' --data 'domainnet'
FedAPT (ours): python prompt-ours.py --logname 'xxxx' --datapath 'xxxx' --data 'domainnet'

Each domain has five client:

PromptFL: python prompt-promptfl-device.py --logname 'xxxx' --datapath 'xxxx' --data 'domainnet' --alpha 0.5
FedAPT (ours): python prompt-ours-device.py --logname 'xxxx' --datapath 'xxxx' --data 'domainnet' --alpha 0.5