Federated Selective Inter-Client Transfer

This code repository contains implementation of the "FedSeIT" model proposed in "Federated Continual Learning for Text Classification via Selective Inter-client Transfer".

This code is based on the implementation of FedWeIT proposed by: Yoon, Jaehong, Wonyong Jeong, Giwoong Lee, Eunho Yang, and Sung Ju Hwang. "Federated continual learning with weighted inter-client transfer." In International Conference on Machine Learning, pp. 12073-12086. PMLR, 2021.

However, we have made a large number of modifications to the codebase provided by the authors of FedWeIT, as cited in our research literature. Our modifications mostly revolved around (1) FedSeIT etc functionality not present in original work (2) Elimination of certain FedWeIT and FedWeIT legacy code which were not pertinent (3) Readability and Flexibility improvements, as well as Documentation.

Setup

 python3.8 -m venv venv
 source venv/bin/activate
 pip3 install -r requirements.txt

Download 'glove.840B.300d.txt' and place it in Resources/

Data Generation

This needs to be called only once, to create tasks from the dataset

cd Config/r8/
bash data.sh

Run FedWeIT setting

cd Config/r8/
bash FedWeIT.sh

Run FedSeIT setting

cd Config/r8/
bash FedSeIT.sh

Note that our codebase currently refers to FedSeIT as FedPrIT

Other Notes

A discussion of important parameters and their meaning can be found in FedWeIT/parser.py