cuequivariance-ops-cu13

What it is and what it does

cuequivariance-ops-cu13 is a Python package that wraps CUDA kernels for equivariant neural network operations. When imported, it loads a precompiled shared library containing GPU kernels optimized for equivariant primitives—mathematical operations that respect symmetries in data. The package itself contains no Python bindings; it acts as a bridge to the underlying CUDA implementation, designed to accelerate equivariant deep learning workloads on NVIDIA GPUs.

The package is part of the broader cuEquivariance ecosystem and is intended for developers building neural networks that exploit equivariance properties. It requires Python 3.10 or later and depends on nvidia-cublas for linear algebra operations, along with tqdm, nvidia-ml-py, and platformdirs for utility functions. Installation is platform-specific, with separate wheels for aarch64 and x86_64 architectures. Users should consult the cuEquivariance documentation for guidance on how to use the kernels within their applications.

Use it for:

• Accelerating equivariant neural networks on NVIDIA GPUs for tasks like 3D object recognition or molecular property prediction.
• Building deep learning models that leverage rotational, translational, or permutation symmetries without manually implementing CUDA kernels.
• Integrating GPU-optimized equivariant operations into larger PyTorch or TensorFlow workflows for scientific computing.
• Developing graph neural networks or point cloud models that require equivariant convolutions or message passing.
• Prototyping symmetry-aware AI models for physics simulations or chemistry applications on GPU hardware.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Provides CUDA kernels for equivariant neural network operations, loaded as a shared library when imported to enable GPU-accelerated equivariant primitives.

Yes, if you are building equivariant neural networks on NVIDIA GPUs and need GPU acceleration. The package is actively maintained, has no known vulnerabilities, and is part of an established NVIDIA ecosystem. However, accept the proprietary license terms carefully: the SDK is restricted to NVIDIA GPU systems, prohibits reverse engineering, and requires your application to have material functionality beyond the SDK itself. Not suitable if you need open-source licensing or non-NVIDIA GPU support.

Install

cuequivariance-ops-cu13 on PyPI

pip install cuequivariance-ops-cu13

uv add cuequivariance-ops-cu13

poetry add cuequivariance-ops-cu13

Installing cuequivariance-ops-cu13

Before you install

Medium install friction due to platform-specific wheel requirements (aarch64 and x86_64 manylinux builds). Active maintenance with a release 7 days ago. Requires Python 3.10 or later and depends on nvidia-cublas, tqdm, nvidia-ml-py, and platformdirs.

License in practice

Licensed under NVIDIA's proprietary SDK agreement with significant restrictions: use limited to systems with NVIDIA GPUs, no reverse engineering or sublicensing, and applications must have material additional functionality beyond the SDK. Distribution requires compliance with specific terms and NVIDIA notification obligations. Liability capped at US$10.00.

Quickstart

pip install cuequivariance-ops-cu13
import cuequivariance_ops_cu13
# Loads CUDA kernels; refer to cuEquivariance documentation for kernel usage

Requires NVIDIA GPU hardware and CUDA-capable system; Python 3.10 or later; nvidia-cublas and other runtime dependencies must be installed.

Verify before relying

• Specific CUDA compute capability requirements and supported GPU architectures not documented in the fact sheet.
• Whether the package works with non-NVIDIA CUDA toolchains or only official NVIDIA CUDA.
• Performance characteristics and typical use-case scale (e.g., model sizes, batch dimensions) not specified.

Package facts

Evidence: cuequivariance_ops_cu13-0.11.1-py3-none-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl; cuequivariance_ops_cu13-0.11.1-py3-none-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl

Tags

More Artificial Intelligence packages

Further reading