DaCeML documentation

Machine learning powered by data-centric parallel programming.

This project adds PyTorch and ONNX model loading support to DaCe, and supports ONNX operators as library nodes to DaCe stateful dataflow multigraphs.

User Guide

• Installation
  • Installing ONNXRuntime
• ONNX
  • Library Nodes
  • Node Implementations
  • Importing ONNX models
  • Schema Representation & Protobuf conversion
• PyTorch Integration
• Automatic Differentiation
  • Using Autodiff
  • Architecture
  • Extending the Engine
• Development
  • Specific Package Versions
  • Makefile Targets
  • Testing
  • Useful Snippets
• Examples

Module Documentation

• daceml.autodiff
  • Generating Backward Passes
  • Extending Autodiff
• daceml.onnx
  • Schema Representation
  • Op Implementation Registration
  • SDFG-based ONNX Implementations
  • Dace CMake Environments
  • Supported ONNX Operators
• daceml.pytorch

Indices and tables

• Index

• Module Index