Welcome to Spox’s documentation!

Spox is a Python framework for constructing ONNX computational graphs.

• The manual serves as a technical explanation of Spox, describing various features.

• The guides serve as a collection of practical examples in the form of Jupyter notebooks.

• Lastly, the generated API listing allows quick access to documentation of public functions.

Manual

• Overview
  • Variables
  • Operator constructors
  • Graphs
  • Running
  • Example usage
• Operators
  • Introduction
  • Representation
  • Usage
• Unstable features
  • Custom operators
  • Functions
  • Value propagation backend
  • Type inference warning levels
  • Initializers
  • Operator overloading

Guides

• Tutorial
  • Introduction
  • Construction
  • Exceptions
  • Building
  • Working with ONNX
  • Summary
• Inline
  • Introduction
  • Composition
  • Custom operators
  • Converter libraries
• Inference
  • Introduction
  • Type inference
  • Value propagation
• Converters
  • Functions
  • Estimators
  • Converter
• Advanced features
  • Introduction
  • Control flow
  • Sequences
  • Example - dynamic piecewise function

API Listing

• API Reference
• Operator constructors for the ai.onnx domain version 17
• Operator constructors for the ai.onnx domain version 18
• Operator constructors for the ai.onnx domain version 19
• Operator constructors for the ai.onnx domain version 20
• Operator constructors for the ai.onnx domain version 21
• Operator constructors for the ai.onnx.ml domain version 3
• Operator constructors for the ai.onnx.ml domain version 4
• Operator constructors for the ai.onnx.ml domain version 5

Indices and tables

• Index

• Module Index

• Search Page