Welcome to FlexFlow’s documentation!

Getting Started

• Overview
• Building from source
• Docker
• Multinode tutorial

FlexFlow Serve

• Serving Overview

FlexFlow Train

• Training Overview
• Training Interface
  • Keras Interface
  • PyTorch Interface
  • ONNX Support
• Training Examples
  • mT5 Model
    • mT5 in PyTorch
    • mT5 in FlexFlow
• Python API
  • Models API
    • Model Creation
      • Model Creation
      • Tensor Creation
    • Model Initialization
      • Compile
      • Initialization
    • Model Training and Testing
      • Fit
      • Evaluate
      • Customized Training
  • Layers API
    • Conv2D
    • Pool2D
    • Dense
    • Embedding
    • Transpose
    • Reverse
    • Concatenate
    • Split
    • Reshape
    • Flat
    • BatchNorm
    • BatchMatMul
    • Add
    • Subtract
    • Multiply
    • Divide
    • Exponential
    • ReLU
    • ELU
    • Sigmoid
    • Tanh
    • Softmax
    • Dropout
    • MultiheadAttention
  • Dataloader API
    • Dataloader Creation
    • Use Dataloader for Training

FlexFlow Backend

• C++ API

Developers Guide

• Developers Guide
  • Code Organization
    • AlexNet example (C++)
  • Continuous Integration
    • Github Workflow syntax
  • Pip packages
    • Packaging
    • Source VS Wheel distribution
    • Versioning
    • Test PyPI
    • Build vs install dependencies
  • Contributing to FlexFlow
    • Formatting
    • Documenting the code
    • Pull Requests
    • Issues
    • License