kernels
Build, optimize, and load fast GPU and CPU compute kernels
All Skills in This Repository (5)
Pure Emerald Level Indicators
huggingfacecuda-kernels
Optimizes NVIDIA GPU AI inference and training with custom CUDA kernels for HuggingFace libraries.
huggingfacecpu-kernels
Writes, optimizes, and benchmarks C++ CPU kernels with AVX2/AVX512 SIMD intrinsics for Hugging Face kernels.
huggingfacexpu-kernels
Generates and optimizes Triton kernels for Intel XPU GPUs through an iterative benchmark-driven workflow.
huggingfacetriton-kernels
Guides writing, testing, and benchmarking portable Triton GPU kernels for NVIDIA and AMD hardware.
huggingfacerocm-kernels
Write and benchmark optimized Triton kernels for AMD GPUs on ROCm.
Frequently Asked Questions
FAQPage SchemaHow to install kernels?▼
Run `npx skills add huggingface/kernels --all -g -y` in your terminal to install all kernel skills globally.
What does the Hugging Face kernels project do?▼
It lets Python apps download and load pre-compiled GPU and CPU kernels from the Hugging Face Hub, so you get fast compute operations without compiling anything locally.
How to write a custom CUDA kernel for PyTorch?▼
The cuda-kernels skill gives your AI agent strict build rules, binding templates, and benchmark scripts to produce compliant kernels for H100, A100, and T4 GPUs.
Can I optimize kernels for AMD or Intel GPUs?▼
Yes. Dedicated skills cover Triton kernels for AMD ROCm GPUs and Intel XPU, including autotuning, profiling, and end-to-end model benchmarks.
Do I need kernel programming experience to use this?▼
No. Your AI agent follows the step-by-step workflows in each skill to write, test, and benchmark kernels from your plain-English request.
Related Repositories in Software Engineering
View All in Software Engineering→
openclawopenclaw
Run a personal AI assistant across your devices and chat apps
obrasuperpowers
Gives coding agents a disciplined workflow from idea to merged code
facebookreact
AI agent skills for building, testing, and porting React core