Welcome to FedML

Thank you for visiting our site. This documentation provides you with everything you need to know about using the FedML platform.

Why FedML?

FedML, Inc. (https://fedml.ai) enables people and/or organizations to have AI capability from data anywhere at any scale. FedML stands for “Fundamental Ecosystem Development/Design for Machine Learning” in a broad scope, and “Federated Machine Learning” in a specific scope. At its current stage, FedML is developing and maintaining a machine learning platform that enables zero-code, lightweight, cross-platform, and provably secure federated learning and analytics. It enables machine learning from decentralized data at various users/silos/edge nodes without requiring data centralization to the cloud, thus providing maximum privacy and efficiency. It consists of a lightweight and cross-platform Edge AI SDK that is deployable over edge GPUs, smartphones, and IoT devices. Furthermore, it also provides a user-friendly MLOps platform to simplify decentralized machine learning and real-world deployment. FedML supports vertical solutions across a broad range of industries (healthcare, finance, insurance, smart cities, IoT, etc.) and applications (computer vision, natural language processing, data mining, and time-series forecasting). Its core technology is backed by over 3 years of cutting-edge research by its co-founders.

Outline

This documentation is organized in the following sections:

Overview

• Getting Started
  • FedML Feature Overview
  • Quick Start for Open Source Library
    • Installation
    • A Quick Overview of the Code Architecture
    • Simplified APIs
  • MLOps User Guide
  • More Resources
    • References
    • Ecosystem
    • Publication
    • Invited Talks (Videos)
    • Join the Community
    • FAQ
    • Join Our Team
• Installing FedML
  • FedML Source Code Repository
  • Installing with pip
  • Installing FedML with Anaconda
  • Installing FedML from Debugging and Editable Mode
  • Installing FedML from Source
  • Running FedML in Docker (Recommended)
  • Running FedML on Kubernetes
• Q&A
  • Guidance for Windows Users
  • Guidance for Raspberry Pi Users
  • Guidance for NVIDIA Jetson Devices
  • Testing if the installation succeeded
  • Installing FedML Android SDK/APP
  • Troubleshooting
• Mission
• Technology Overview
  • 1. Industrial-level System Design and Optimization
  • 2. User-friendly MLOps platform to simplify collaboration and real-world deployment
  • 3. A Full-stack of Scientific Publications in ML Algorithms, Security/Privacy, Systems, Applications, and Visionary Impacts
• Application Ecosystem
  • FedNLP: Benchmarking Federated Learning Methods for Natural Language Processing Tasks
  • FedGraphNN: A Federated Learning Benchmark System for Graph Neural Networks
  • FedCV: A Federated Learning Framework for Diverse Computer Vision Tasks
  • FedIoT: Federated Learning for Internet of Things
• Open Source Code Architecture
  • Code Architecture
  • Reference
  • Video
• FedML MLOps Video Tutorial
• FedML FAQ

FedML MLOps - Landing FedML into Reality

• FedML MLOps Video Tutorial
• FedML MLOps User Guide
  • How Does FedML MLOps Platform Work?
  • Local Development and Building MLOps Packages
  • 1. Create Application and Upload Local Packages
  • 2. Install FedML Agent: fedml login $account_id
  • 3. Invite Collaborators, Create a Group and a Project
  • 4. Start Run (Training) via Automated Deployment and Scaling
  • 5. Experimental Tracking via Simplified Project Management
  • 6. FedML OTA(Over-the-Air) upgrade mechanism
• FedML MLOps CLI and API Reference
  • Overview
  • 1. Login to the FedML MLOps platform (open.fedml.ai)
    • 1.1. Examples for Logging in to the FedML MLOps platform (open.fedml.ai)
  • 2. Build the client and server package in the FedML MLOps platform (open.fedml.ai)
    • 2.1. Examples for building the client and server package
  • 3. Log out the MLOps platform (open.fedml.ai)
  • 4. Display FedML Environment and Version
  • 5. Display logs
  • 6. Diagnosis
  • 7. Jobs

FedML Parrot - Simulating the Real World

• FedML Parrot User Guide
  • Outline
  • Installation
  • Algorithm Reference Implementation
  • Datasets and Models
  • Customizing Data Loader for Your Own Dataset
  • Benchmark
• Config MLOps parameters
• Login to MLOps
  • Examples
• FedML Parrot Examples
  • Simulation with a Single Process (Standalone)
  • Simulation with Message Passing Interface (MPI)
  • Simulation with NCCL-based MPI (the fastest training)
• FedML Parrot API Reference

FedML Octopus - Simple Connector for Data Silos

• FedML Octopus User Guide
  • Seamlessly transplant the simulation code (Parrot) to real-world cross-device FL (Octopus)
  • Heterogeneous hierarchical federated learning: supporting local AllReduce-based distributed training
  • Diverse Communication Backends for different cross-silo scenario
• FedML Octopus Examples
  • Horizontal Federated Learning
  • Hierarchical Federated Learning
• FedML Octopus API Reference

FedML BeeHive - Collaborative Learning on Smartphones/IoTs

• FedML Beehive User Guide
• FedML Beehive Examples
• FedML BeeHive API Reference

FedML Cheetah - Speedy Training of Large Models

• FedML Cheetah User Guide

Benchmarks for FedNLP, FedCV, FedGraphNN and FedIoT

• Federated Graph Neural Networks (FedGraphNN)
  • Graph-Level Federated Graph Neural Networks
  • Subgraph-Level Federated Graph Neural Networks
  • Node-Level Federated Graph Neural Networks
  • Federated OGB and OGB-LSC Datasets
    • Challenges and Leaderboards

Resources

• Research Publications
  • Outline
  • Vision Paper for High Scientific Impacts
  • System for Large-scale Distributed/Federated Training
  • Training Algorithms for FL
  • Security/privacy for FL
  • AI Applications
• Video
• Join the Community
  • Join our Slack:
  • Join our Discord:
  • Join our WeChat Group:

Careers

FedML is hiring researchers, engineers, product managers, and related interns. If you are interested, Please apply at https://fedml.ai/careers