flwr-nightly 0.1.1.dev20200713

Flower - A Friendly Federated Learning Research Framework

Flower - A Friendly Federated Learning Research Framework

Flower is a research framework for building federated learning systems. The design of Flower is based on a few guiding principles:

• Customizable: Federated learning systems vary wildly from one use case to another. Flower allows for a wide range of different configurations depending on the needs of each individual use case.

• Extendable: Flower originated from a research project at the Univerity of Oxford, so it was build with AI research in mind. Many components can be extended and overridden to build new state-of-the-art systems.

• Framework-agnostic: Different machine learning frameworks have different strengths. Flower can be used with any machine learning framework, for example, PyTorch, TensorFlow, or even raw NumPy for users who enjoy computing gradients by hand.

• Understandable: Flower is written with maintainability in mind. The community is encouraged to both read and contribute to the codebase.

Note: Even though Flower is used in production, it is published as pre-release software. Incompatible API changes are possible.

Installation

Flower can be installed directly from the GitHub repository using pip:

$ pip install git+https://github.com/adap/flower.git

Official PyPI releases will follow once the API matures.

Run Examples

We built a number of examples showcasing different usage scenarios in src/flower_example. To run an example, first install the necessary extras (available extras: examples-tensorflow):

pip install git+https://github.com/adap/flower.git#egg=flower[examples-tensorflow]

Once the necessary extras (e.g., TensorFlow) are installed, you might want to run the Fashion-MNIST example by starting a single server and multiple clients in two terminals using the following commands.

Start server in the first terminal:

$ ./src/flower_example/tf_fashion_mnist/run-server.sh

Start the clients in a second terminal:

$ ./src/flower_example/tf_fashion_mnist/run-clients.sh

Docker

If you have Docker on your machine you might want to skip most of the setup and try out the example using the following commands:

# Create docker network `flower` so that containers can reach each other by name
$ docker network create flower
# Build the Flower docker containers
$ ./dev/docker_build.sh

# Run the docker containers (will tail a logfile created by a central logserver)
$ ./src/flower_example/tf_fashion_mnist/run-docker.sh

This will start a slightly reduced setup with only four clients.

Documentation

• Documentation

Contributing to Flower

We welcome contributions. Please see CONTRIBUTING.md to get started!