xain-fl 0.6.0

XAIN is an open source framework for federated learning.

---

Disclaimer: This is work-in-progress and not production-ready, expect errors to occur! Use at your own risk! Do not use for any security related issues!

---

XAIN

The XAIN project is building a privacy layer for machine learning so that AI projects can meet compliance such as GDPR and CCPA. The approach relies on Federated Learning as enabling technology that allows production AI applications to be fully privacy compliant.

Federated Learning also enables different use-cases that are not strictly privacy related such as connecting data lakes, reaching higher model performance in unbalanced datasets and utilising AI models on the edge.

This repository contains the source code for running the Coordinator. The Coordinator is the component of Federated Learning that selects the Participants for training and aggregates the models using federated averaging.

The Participants run in a separate environment than the Coordinator and connect to it using an SDK. You can find here the source code for it.

Quick Start

XAIN requires Python 3.6.4+. To install the xain-fl package just run:

$ python -m pip install xain-fl

Install from source

Clone this repository:

git clone https://github.com/xainag/xain-fl.git

Install this project with the dev profile (NOTE: it is recommended to install the project in a virtual environment):

cd xain-fl
pip install -e '.[dev]'

Verify the installation by running the tests

pytest

Building the Documentation

The project documentation resides under docs/. To build the documentation run:

$ cd docs/
$ make docs

The generated documentation will be under docs/_build/html/. You can open the root of the documentation by opening docs/_build/html/index.html on your favorite browser or simply run the command:

$ make show

Running the Coordinator locally

To run the Coordinator on your local machine, you can use the example-config.toml file:

# If you have installed the xain_fl package,
# the `coordinator` command should be directly available
coordinator --config configs/example-config.toml

# otherwise the coordinator can be started by executing the
# `xain_fl` package:
python xain_fl --config configs/example-config.toml

Run the Coordinator from a Docker image

There are two docker-compose files, one for development and one for release.

Development image

To run the coordinator's development image, first build the Docker image:

$ docker build -t xain-fl-dev -f Dockerfile.dev .

Then run the image, mounting the directory as a Docker volume:

$ docker run -v $(pwd):/app -v '/app/xain_fl.egg-info' xain-fl-dev coordinator

The command above uses a default configuration but you can also use a custom config file:

For instance, if you have a ./custom_config.toml file that you'd like to use, you can mount it in the container and run the coordinator with:

docker run \
  -v $(pwd)/custom_config.toml:/custom_config.toml \
  -v $(pwd):/app \
  -v '/app/xain_fl.egg-info' \
  xain-fl-dev \
  coordinator --config /custom_config.toml

Release image

To run the coordinator's release image, first build it:

$ docker build -t xain-fl .

And then run it (this example assumes you'll want to use the default port):

$ docker run -p 50051:50051 xain-fl

Docker-compose

The coordinator needs a storage service that provides an AWS S3 API. For development, we use minio. We provide docker-compose files that start coordinator container along with a minio container, and pre-populate the appropriate storage buckets.

Development

To start both the coordinator and the minio service use:

docker-compose -f docker-compose-dev.yml up

It is also possible to only start the storage service:

docker-compose -f docker-compose-dev.yml up minio-dev initial-buckets

Release

$ docker-compose up

Related Papers and Articles

• An introduction to XAIN’s GDPR-compliance Layer for Machine Learning
• Communication-Efficient Learning of Deep Networks from Decentralized Data
• Analyzing Federated Learning through an Adversarial Lens
• Towards Federated Learning at Scale: System Design