A Domain Specific Language for Federated Learning

A Domain-Specific Language for Federated Learning

The FastFL Python interface is now extended with the possibility to specify FL workloads through a DSL, abstracting the ML practitioner from the low-level implementation details of the federation.

This release currently supports just the master-worker scenario training on MNIST on clusters with a shared file system.

v0.1.0-alpha

First public release of FFL - Fast Federated Learning.

This release contains three different examples of use of FFL for Federated Learning (FL) and Edge inference (EI):

1. Master-worker (FL)
2. Peer-to-peer (FL)
3. Tree-based (EI)

The setup.sh script downloads and installs all the necessary software for running the experiments (apart from GCC, make, unzip, and OpenCV, which are on the user).
The reproduce.sh script is designed to reproduce the full suite of experiments reported in the "Experimenting with Emerging RISC-V Systems for Decentralised Machine Learning", published at the ACM Computing Frontiers 2023 conference.

Full Changelog: https://github.com/alpha-unito/FastFederatedLearning/commits/v0.1.0-alpha