Hugging Face has detailed major updates to its 🤗 Kernels project, which aims to standardize how custom kernels are packaged, distributed, and consumed securely. Over the past few months, the project has been almost completely redesigned. A new repository type called 'kernel' has been introduced on the Hub, allowing users to browse available kernels and see supported accelerators, operating systems, and backend versions. To address security concerns regarding native code execution, the project incorporates Nix for reproducible builds, embeds source Git SHA1 for provenance, and implements trusted kernel publishers. By default, the kernels package will only load kernels from trusted publishers, requiring explicit opt-in for others. Additional security features include code signing using Sigstore's cosign with ephemeral private keys, though full signature verification upon loading is still being tested. Furthermore, Hugging Face has established a clearer separation of concerns between the kernels library and kernel-builder, creating a leaner command line interface tailored for both human users and the rise of agentic kernel development. Agents can now leverage these tools to scaffold, build, benchmark, and iteratively optimize kernels against defined baselines using integrations like HF Jobs. To simplify setup, the project now offers a one-click installation script and a Terraform setup guide, alongside system cards to expose useful kernel interface information.
🤗 Kernels: Major Updates
Hugging Face has announced major updates to its 🤗 Kernels project, featuring a complete redesign focused on packaging, distributing, and consuming custom kernels securely. The updates introduce a new Hub repository type, enhanced security measures including trusted publishers and code signing, and better support for agentic kernel development.
Why it matters
Standardizing and securing custom kernels on the Hugging Face Hub makes high-performance compute components more discoverable and reliable. By supporting agentic kernel development and reproducible builds, these updates streamline optimization workflows across diverse hardware.
Topic dossiers Hugging Face