Federated Learning (FL) development kit for FedCampus.
A FatKid to support the FatCampus.
This repository contains the libraries used to implement the FL functionalities of the FedCampus app.
The backend server uses Django REST Framework to provide persistent service and on-demand, backend-configurable training.
The FL training steps are supported using the Flower FL framework. Flower servers use gRPC, and are spawned on-demand by the backend server.
Located at backend/
. Please see backend/README.md
for information on how to set up the backend server.
The Android client package and demo is located at client/
.
To try out the Android client, please see client/README.md
.
To use the library in this repository to implement custom Android Flower clients, please see client/fedcampus/README.md
.
The Flutter client package to communicate with the backend and Flower servers is at flutter/
.
The example Flutter client for both Android and iOS is at fed_kit_client/
. To try it out, please see fed_kit_client/README.md
.
We support TensorFlow (Keras) models for Android and Core ML models for iOS.
We also include an example that converts a TensorFlow model with mean square error to Core ML using coremltools
.
The ML model generation script for Android is located at gen_tflite/
. Please see gen_tflite/README.md
for information on how to create models and convert them to .tflite
files.
gen_mlmodel
is still a work in progress. Meanwhile, you can construct Core ML models manually using coremltools
.
data_type
.requirements.txt
s.Please see CONTRIBUTING.md for FedCampus.
This repository is moved from dyn_flower_android_drf to FedKit. Please see the older Git history in the former repository.