cuequivariance

Construct group-equivariant neural network layers using irreducible representations and segmented tensor products.

413|62|Updated Jun 23, 2026

Search Tags:#group theory #geometry #neural networks #equivariant #cuequivariance #tensor products

Ready to run

npx skills add https://github.com/NVIDIA-BioNeMo/bionemo-agent-toolkit --skill cuequivariance-nvidia-bionemo

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: cuequivariance
Source: https://github.com/NVIDIA-BioNeMo/bionemo-agent-toolkit/tree/main/plugins/bionemo-agent-toolkit/skills/cuEquivariance
Command: npx skills add https://github.com/NVIDIA-BioNeMo/bionemo-agent-toolkit --skill cuequivariance-nvidia-bionemo

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires numpy.

What problem does it solve?

This skill addresses the complexity of implementing group-equivariant neural networks by providing a robust framework for managing irreducible representations and segmented tensor products.

Core Features & Use Cases

Group Theory Abstractions: Define custom Lie groups (SO3, O3, SU2) and manage irreducible representations with multiplicities.
Segmented Polynomials: Construct complex equivariant operations using segmented tensor products and Clebsch-Gordan coefficients.
Use Case: Researchers building geometric deep learning models for molecular structure prediction can use this to define custom equivariant layers that respect the physical symmetries of the system.

Quick Start

Use the cuequivariance skill to define a custom SO3 representation and build a fully connected tensor product for a geometric neural network layer.

Frequently Asked Questions about cuequivariance

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I build equivariant neural network layers for molecular modeling?▼

To build equivariant neural network layers, use this framework to manage irreducible representations and segmented tensor products. It constructs operations respecting SO3, O3, and SU2 symmetry constraints for geometric deep learning in molecular and physical modeling.

What are irreducible representations in geometric deep learning?▼

Irreducible representations are mathematical abstractions used to define custom Lie groups like SO3 and SU2. They manage symmetry constraints with multiplicities to construct complex equivariant operations using Clebsch-Gordan coefficients and segmented tensor products.

How do I implement Clebsch-Gordan coefficients for tensor products in neural networks?▼

Implement Clebsch-Gordan coefficients for tensor products by utilizing the segmented polynomial generation features. This framework computes high-performance tensor contractions and descriptor-based equivariant polynomials for custom geometric neural network layers.

Can I use numpy to define custom SO3 representations for scientific machine learning?▼

Yes, you can define custom SO3 representations using numpy, which is the required dependency. This allows you to build fully connected tensor products and manage group theory abstractions for scientific machine learning tasks.

When do I need group-equivariant neural networks for molecular structure prediction?▼

You need group-equivariant neural networks when predicting molecular structures requires respecting physical symmetries. This framework provides the necessary group theory abstractions and tensor contractions to enforce these constraints during model training.