Kernels

The Kernel Hub allows Python libraries and applications to load compute kernels directly from the Hub. Kernels are a first-class repository type on the Hub, with dedicated pages that surface supported hardware and versions. To support dynamic loading, Hub kernels differ from traditional Python kernel packages in that they are made to be:

• Portable: a kernel can be loaded from paths outside PYTHONPATH.
• Unique: multiple versions of the same kernel can be loaded in the same Python process.
• Compatible: kernels must support all recent versions of Python and the different PyTorch build configurations (various CUDA versions and C++ ABIs). Furthermore, older C library versions must be supported.

Browse available kernels at huggingface.co/kernels.

The Kernels project is divided into two parts:

• Builder: kernel-builder provides utilities to build, package, and distribute compute kernels in a way that is compatible with the Hugging Face Hub and kernels.
• kernels: The kernels is a Python package that lets users load compatible compute kernels from the Hub. Refer to the quickstart to know more.

If you’re looking for a more involved “Why kernels?” answer, refer to this page.

The talks page page has links to talks on the Kernels project.