Tensor Comprehensions documentation

Tensor Comprehensions provides framework-agnostic abstractions for High-Performance Machine Learning.

Index

• What is Tensor Comprehensions?
• Semantics
• Range Inference
• Relation to Halide
• Mapping Options
• Autotuner
• Performance of TC

Machine Learning with TC

• Positioning of TC in ML Software stacks
• Integrating TC with ML framework
• Coding Conventions

PyTorch Integration

• Getting Started
• Python API
• Writing TC operations
• Autograd with TC
• Debugging
• Frequently Asked Questions

Installation

• Installation Guide
• Installation in the Google Colaboratory environment

Paper

• Tech Report

Support

• Contacts