User Guide

This user guide provides instructions on how to utilize various features in NVIDIA FLARE.

For a more in-depth exploration of the capabilities offered by apps with custom workflows and algorithms, please refer to the Programming Guide.

Integration Developers: Developers who works on building Platform powered by NVIDIA FLARE. refer to Programming Guide for APIs needed for integration

Data Scientists: Users that focused on federated learning applications. They are less interested in develop custom algorithms or workflow patterns, applying the existing FL algorithm to real-world applications is their focus. please check Data Scientist Guide

Researchers: Users who focused on FL research, needs to experiments and customize both client and server code to develop new workflow. please check Researcher Guide

• NVIDIA FLARE Application
  • FL server configuration
  • FL client configuration
  • Custom code
    • Custom code config example
    • Troubleshooting BYOC
  • Resources
• Jobs: Defining Jobs and Multi-Job Execution
  • Concepts
    • Job
    • Resources
    • Deploy Map
    • Resource-less Jobs
    • Resource-Based Job Automation
    • Job Runner
    • Job Scheduler Configuration
• NVIDIA FLARE Workspace
  • Server
    • Accessing server-side workspace
  • Client
• Admin Guide
  • Configuration Files
    • Variable Resolution in Job Configuration
    • Predefined Job Configuration Variables
    • FL Server Port Consolidation
    • Communication Configuration
    • Logging Configuration
  • Deployment Guide
    • Deployment Overview
    • Operating NVFLARE - Admin Client, Commands, FLARE API
    • Containerized Deployment
    • Helm Chart for NVIDIA FLARE
    • Cloud Deployment
    • Amazon Elastic Kubernetes Service Deployment
    • NVFLARE Dashboard UI
    • Large Models
  • NVIDIA FLARE Security
    • Security Framework
  • Monitoring
    • Overview
    • Key Components
    • Configuration
    • Visualization
    • Summary
• Data Scientist Guide
  • Client API Usage
    • Core concept
  • NVFlare Job Recipe
    • Motivation for Using JobRecipe
    • Basic Example
    • Execution Environments
    • SimEnv – Simulation Environment
    • PocEnv – Proof-of-Concept Environment
    • ProdEnv – Production Environment
    • Benefits of Environment Abstraction
    • Special Considerations for Edge Applications
    • Best Practices
    • Summary
    • Examples
  • FLARE API
    • Initialization and Usage
    • Implementation Notes
    • Additional and Complex Commands
  • POC: Prove of Concept: Simulate Production deployment locally
    • Setting Up the Application Environment in POC Mode
    • Starting the Application Environment in POC Mode
  • Integration of Flower Applications with NVIDIA FLARE
    • Initial Integration
    • Flower Job Structure
    • Run Flower Application as FLARE Job
    • FLARE Multi-Job Architecture
    • Detailed Design
    • Reliable Messaging
  • Federated XGBoost with NVFlare
    • NVFlare XGBoost User Guide
    • Reliable Federated XGBoost Design
    • Reliable Federated XGBoost Timeout Mechanism
    • Secure Federated XGBoost Design
• Researcher Guide
  • Researcher Area
    • FedSM
    • Auto-FedRL
    • Quantifying Data Leakage in Federated Learning
• NVFlare CLI
  • NVIDIA FLARE FL Simulator
    • Command Usage
    • Command examples
    • Debug NVFlare Application
    • Processes, Clients, and Events
  • Proof Of Concept (POC) Command
    • Syntax and Usage
    • Set Up POC Workspace
    • Start Package(s)
    • Operating the System and Submitting a Job
    • Stop Package(s)
    • Clean Up
    • Learn More
  • Provision command
    • Dynamic Provisioning
  • NVIDIA FLARE Job CLI
    • Command Usage
    • Command examples
    • Advanced Job Configurations
  • NVIDIA FLARE Preflight Check
    • General Usage
    • Preflight check on server site
    • Preflight check on client sites
    • Preflight check for admin consoles
  • Dashboard Command
    • Introduction to the Dashboard Command
    • Syntax and Usage
  • NVFLARE Code Pre-Installer
    • Overview
    • Directory Structure
    • Usage
    • Application Code Structure
    • Installation Paths
    • Error Handling
    • Notes
    • Using Pre-installed Code when submit job
• FLARE Confidential Federated AI
  • Introduction
    • Security Risks in Federated Learning
    • What is Confidential Computing?
    • Risk Mitigation with Confidential Computing
    • Operational Risks Even with Confidential Computing
  • FLARE Confidential Federated AI Overview
    • On-Premises IP Protection Deployment
    • Azure Confidential Computing Deployment
  • Choosing the Right Deployment
    • On-Premises IP Protection
    • Azure Confidential Computing Deployment
• Develop Edge Applications with FLARE
  • Deployment Architecture
  • Hierarchical FLARE
  • Edge Training Algorithm
    • Flexible Orchestration Patterns from Synchronous to Asynchronous Aggregation
  • Edge Device Interaction Protocol (EDIP)
    • Step 1 - Get a Job
    • Step 2 - Get a Task
    • Step 3 - Execute Task and Report Result
  • Device Simulation
    • Simulation Logic
    • Step 1 - Get a Job
    • Step 2 - Get Selections
    • Step 3 - Get and Execute Task
    • Simulation Completion
  • Simulator Configuration
  • How to run simulation
  • Model Development
    • Step 1 - Design Model Architecture
    • Step 2 - Create DeviceModel
    • Step 3 - Create FLARE Job
    • EdgeFedBuffRecipe
  • Federated Training on Mobile Devices
    • ETFedBuffRecipe