This repo contains a Nix package that can be used to build custom machine learning kernels for PyTorch. The kernels are built using the PyTorch C++ Frontend and can be loaded from the Hub with the kernels Python package.

This builder is a core component of the larger kernel build/distribution system.

Torch 2.9 note: kernel-builder currently builds Torch 2.9 extensions based on the fifth release candidate. If you upload Torch 2.9 kernels, please validate them against the final Torch 2.9.0 release. In the unlikely case of an ABI-breaking change, you can rebuild and upload your kernel once kernel-builder is updated for the final release

We recommend using Nix to build kernels. To speed up builds, first enable the Hugging Face binary cache:

# Install cachix and configure the cache
cachix use huggingface

# Or run once without installing cachix
nix run nixpkgs#cachix -- use huggingface

Then quick start a build with:

cd examples/activation
nix run .#build-and-copy \
  --override-input kernel-builder github:huggingface/kernel-builder \
  --max-jobs 8 \
  -j 8 \
  -L

The compiled kernel will then be available in the local build/ directory. We also provide Docker containers for CI builds. For a quick build:

# Using the prebuilt container
cd examples/activation
docker run --rm \
  --mount type=bind,source=$(pwd),target=/kernelcode \
  -w /kernelcode ghcr.io/huggingface/kernel-builder:main build

See dockerfiles/README.md for more options, including a user-level container for CI/CD environments.

• Writing Hub kernels
• Building kernels with Nix
• Building kernels with Docker (for systems without Nix)
• Local kernel development (IDE integration)
• Kernel security
• Why Nix?

The generated CMake build files are based on the vLLM build infrastructure.