---
id: kernels-data
version: "0.16.0"
license: unclear
license_treatment: unclear
maintenance: active
---

# kernels-data — Kernels data structures (Python bindings)

License: unclear · Maintenance: active · Downloads: 975.5K/mo

## What it is and what it does

kernels-data is a compiled Python package providing data structures and bindings for the Hugging Face kernels ecosystem. It enables dynamic loading of optimized compute kernels from the Hugging Face Hub, supporting portable, versioned kernel execution across different PyTorch and CUDA configurations. The package is built with Rust and distributed as platform-specific wheels covering CPython 3.8+ on Linux, macOS, and Windows architectures.

The package is designed as part of a larger kernel-loading system where kernels can be fetched from the Hub and executed without modifying PYTHONPATH or dealing with traditional Python packaging constraints. It abstracts away compatibility concerns across different build variants and older C library versions, making it a foundational piece for applications that need to run specialized compute kernels on GPUs.

Use it for:

- Load and execute optimized CUDA kernels from the Hugging Face Hub in GPU-accelerated applications.
- Build inference pipelines that dynamically fetch and run specialized compute kernels without recompilation.
- Support multiple kernel versions in the same Python process for testing or gradual migration of kernel implementations.
- Integrate with applications that rely on custom optimized kernels for performance-critical operations.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

kernels-data provides Python bindings for kernel data structures used by the Hugging Face kernels ecosystem, enabling dynamic loading and execution of optimized compute kernels from the Hub.

Yes, if you are building applications that use Hugging Face kernels or models requiring dynamic kernel loading from the Hub. The package is actively maintained, has no known vulnerabilities, and is widely used (top 5000 on PyPI with 975470 monthly downloads). However, verify the unclear license terms before use in proprietary projects, and confirm your environment meets the stated requirements.

## Install

pip install kernels-data
uv add kernels-data
poetry add kernels-data

## Installing kernels-data

Before you install: Medium install friction due to platform-specific wheels and compiled bindings (Rust). Active maintenance with recent releases and no known vulnerabilities.

License in practice: License treatment is unclear—no SPDX identifier or raw license text provided. Verify the actual license before using in proprietary or restricted-license projects.

Quickstart:

pip install kernels-data

from kernels import get_kernel

kernel = get_kernel("kernels-community/activation", version=1)

Requires torch>=2.5 and CUDA. Python >=3.8 required.

Verify before relying:

- Exact license identifier and terms—license_treatment is marked unclear with no SPDX or raw text provided.
- Whether kernels-data is a standalone package or a data/binding component that requires additional packages for practical use.
- Compatibility matrix for specific PyTorch and CUDA versions beyond the stated torch>=2.5 requirement.

## Package facts

- License: not declared (unclear)
- Python support: supports_current
- Install friction: medium
- Maintenance: active
- Downloads: 975.5K/month (top 5,000 on PyPI)
- Known vulnerabilities: none known

## Tags

hugging face kernel data structures, dynamic compute kernel execution, optimized kernel bindings python, hub kernel data structures, kernel management hugging face, cuda kernel hub integration, portable kernel loading, gpu-acceleration, hugging-face, pytorch-kernels

[View on SkillFed](https://skillfed.io/packages/kernels-data) · [View on PyPI](https://pypi.org/project/kernels-data/)