Flower A Friendly Federated Learning Framework

A unified approach to federated learning, analytics, and evaluation. Federate any workload, any ML framework, and any programming language.

Take the tutorial

to learn federated learning

Users love Flower

“In just a few lines of code I was able to federate my Machine Learning project.”

Sandra Carrasco Limeros, Deep Learning Researcher at AI Sweden

“The flexibility of Flower is precisely what we need to develop production-ready solutions for our customers.”

Jan Schlicht, Lead Engineer at Katulu

“Scaling FL experiments to 1,000 or even 10,000 clients can be very challenging. Flower takes care of all the scaling complexities and allows researchers like me to focus on writing client and server-side algorithms.”

Akhil Mathur, Principal Research Scientist at Nokia Bell Labs

“I love working with Flower because it is so easy to use and very compatible with, e.g., PyTorch, NumPy, and Weights & Biases.”

Xinchi Qiu, PhD Student at the University of Cambridge

“Flower was a natural choice for us [...]: it is agnostic to any particular ML framework, [...]; in addition, it allows both to run simulations on a single machine and to develop real FL systems ready to be deployed, almost using the same code.”

Paolo Bellavista, Professor at the University of Bologna

“My Students are free to choose their FL Framework, but Flower has proven itself because it’s difficult to get wrong!”

Afra Mashhadi, Assistant Professor at the University of Washington

“Flower was much much easier to implement. We were wishing we had started with Flower to begin with!”

Jason Mixon, PhD Student and Instructor at Dakota State University

“Federated learning has always been a concept that has been daunting to me. It only took me 3 hours to understand it with Flower, using the great documentation and tutorials.”

James Faure, Student Researcher at Stellenbosch University

Some of the best organizations in the world use Flower

Federate easily

Getting started with Flower can be so easy. Code examples show different usage scenarios of Flower in combination with popular machine learning frameworks.

0. Install Flower and TensorFlow

pip install flwr tensorflow

1. client.py

import flwr as fl
import tensorflow as tf

# Load model and data (MobileNetV2, CIFAR-10)
model = tf.keras.applications.MobileNetV2((32, 32, 3), classes=10, weights=None)
model.compile("adam", "sparse_categorical_crossentropy", metrics=["accuracy"])
(x_train, y_train), (x_test, y_test) = tf.keras.datasets.cifar10.load_data()

# Define Flower client
class CifarClient(fl.client.NumPyClient):
  def get_parameters(self, config):
    return model.get_weights()

  def fit(self, parameters, config):
    model.set_weights(parameters)
    model.fit(x_train, y_train, epochs=1, batch_size=32)
    return model.get_weights(), len(x_train), {}

  def evaluate(self, parameters, config):
    model.set_weights(parameters)
    loss, accuracy = model.evaluate(x_test, y_test)
    return loss, len(x_test), {"accuracy": accuracy}

# Start Flower client
fl.client.start_numpy_client(server_address="127.0.0.1:8080", client=CifarClient())

2. server.py

import flwr as fl

# Start Flower server
fl.server.start_server(
  server_address="0.0.0.0:8080",
  config=fl.server.ServerConfig(num_rounds=3),
)

Getting Started

Installation Guide

The Flower documentation has detailed instructions on what you need to install Flower and how you install it. Spoiler alert: you only need pip! Check out our installation guide.

PyTorch, TensorFlow, 🤗, ...?

Do you use PyTorch, TensorFlow, scikit-learn, MXNet, or Hugging Face? Then simply follow our quickstart examples that help you to federate your existing ML projects.

_Why Flower?

A unified approach to federated learning, analytics, and evaluation.

Scalability

Flower was built to enable real-world systems with a large number of clients. Researchers used Flower to run workloads with tens of millions of clients.

ML Framework Agnostic

Flower is compatible with most existing and future machine learning frameworks. You love Keras? Great. You prefer PyTorch? Awesome. Raw NumPy, no automatic differentiation? You rock!

Cloud, Mobile, Edge & Beyond

Flower enables research on all kinds of servers and devices, including mobile. AWS, GCP, Azure, Android, iOS, Raspberry Pi, Nvidia Jetson, all compatible with Flower.

Research to Production

Flower enables ideas to start as research projects and then gradually move towards production deployment with low engineering effort and proven infrastructure.

Platform Independent

Flower is interoperable with different operating systems and hardware platforms to work well in heterogeneous edge device environments.

Usability

It's easy to get started. 20 lines of Python is enough to build a full federated learning system. Check the code examples to get started with your favorite framework.

Join ourCommunity!

Join us on our journey to make federated approaches available to everyone.

Give feedback

Join Slack