Does kernels auto-invoke its skills?

Not yet. It is indexed on Flowy as a plain plugin. Request auto-invocation on its page and Flowy will route its skills for you as you prompt.

Yes. Flowy is free and open source, with nothing gated. You can read every skill in full before you install.

Development

Plugin

kernels

The Kernel Hub allows Python libraries and applications to load compute kernels directly from the Hub.

FLOW.md

huggingface@huggingface

GitHub

Plugin Signal70/100Solid

maintained99

capable37

documented75

adopted61

Install

$ npx -y skills add huggingface/kernels --agent claude-code

Run the curl in your terminal, the rest in Claude Code.

Stats

Python Bash C++GitHub Hugging Face

723

Stars

119

Forks

Active

Maintenance

Python

Language

Apache-2.0

License

2h ago

Last commit

1y ago

Created

Repo: huggingface/kernels

In this plugin

5 skills

See everything inside

What's inside

Skills

/cpu-kernelsProvides guidance for writing, optimizing, and benchmarking C++ CPU kernels with SIMD intrinsics (AVX2/AVX512) for the Hugging Face kernels ecosystem. Includes a two-phase workflow: Phase 1 correctness (generic → AVX2) and Phase 2 performance exploration (AVX512 with branching /cuda-kernelsProvides guidance for writing and benchmarking optimized CUDA kernels for NVIDIA GPUs (H100, A100, T4) targeting HuggingFace diffusers and transformers libraries. Kernels must be kernel-builder/ABI3-compliant: no pybind11, no setup.py, TORCH_LIBRARY_EXPAND bindings only./rocm-kernelsProvides guidance for writing and benchmarking optimized Triton kernels for AMD GPUs (MI355X, R9700) on ROCm, targeting HuggingFace diffusers (LTX-Video, SD3, FLUX) and transformers. Core kernels: RMSNorm, RoPE 3D, GEGLU, AdaLN. Includes XCD swizzle, autotune, diffusers /triton-kernels| name | triton-kernels | | --- | --- | | description | Provides guidance for writing and benchmarking portable Triton kernels targeting NVIDIA and AMD GPUs. Covers core DSL patterns, @triton.autotune, numerics (fp16/bf16/fp8), masked loads, reductions, tiling, benchmarking /xpu-kernelsProvides guidance for writing, optimizing, and benchmarking Triton kernels for Intel XPU GPUs (Battlemage/Arc Pro B50) using the Xe-Forge optimization framework. Includes an LLM-driven trial-loop workflow (analyze, validate, benchmark, profile, finalize), XPU-specific patterns

Badges

README

Read on GitHub

kernels

The Kernel Hub allows Python libraries and applications to load compute kernels directly from the Hub. To support this kind of 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.

Components

You can load kernels from the Hub using the kernels Python package.
If you are a kernel author, you can build your kernels with kernel-builder.
Hugging Face maintains a set of kernels in kernels-community.

🚀 Quick Start

Install the kernels Python package with pip (requires torch>=2.5 and CUDA):

pip install kernels

Here is how you would use the activation kernels from the Hugging Face Hub:

import torch

from kernels import get_kernel

# Download optimized kernels from the Hugging Face hub
activation = get_kernel("kernels-community/activation", version=1)

# Random tensor
x = torch.randn((10, 10), dtype=torch.float16, device="cuda")

# Run the kernel
y = torch.empty_like(x)
activation.gelu_fast(y, x)

print(y)

Browse available kernels at huggingface.co/kernels.

📚 Documentation

Read the documentation of kernels and kernel-builder.

File tree

727

huggingface/kernelsView on GitHub

.github/
  dependabot.yml
  ISSUE_TEMPLATE/
    bug_report.yml
    feature_request.yml
  pull_request_template.md
  VOUCHED.td
  workflows/
    build_documentation.yaml
    build_kernel_cpu.yaml
    build_kernel_macos.yaml
    build_kernel_rocm.yaml
    build_kernel_windows.yaml
    build_kernel_xpu.yaml
    build_kernel.yaml
    build_pr_documentation.yaml
    check_variants.yaml
    coverage_comment.yaml
    kernel-builder-cli-docs.yaml
    kernels_data_python_release.yaml
    lint.yml
    nix_checks.yml
    publish_kernels.yml
    rust.yaml
    security-audit.yml
    test_e2e.yaml
    test_extra_commands.yaml
    test_kernels.yaml
    test_python.yaml
    update_abi_symbols.yaml
    update_cache.yaml
    upload_pr_documentation.yaml
    vouch-check-pr.yaml
.gitignore
.pre-commit-config.yaml
Cargo.lock
Cargo.toml
CONTRIBUTING.md
default.nix
docs/
  source/
    _toctree.yml
    api/
      kernels.md
      layers.md
    basic-usage.md
    blog.md
    builder/
    builder-cli.md
      agents-guide.md
      build-variants.md
      build.md
      design-nix-builder.md
      design-overview.md
      github-actions.md
      ide-setup.md
      local-dev.md
      metal.md
      security.md
      why-nix.md
      writing-kernels.md
    cli-benchmark.md
    cli-download.md
    cli-info.md
    cli-lock.md
    cli-skills.md
    cli-verify-signature.md
    cli-versions.md
    cli.md
    env.md
    faq.md
    index.md
    installation.md
    integrating-kernels.md
    kernel-requirements.md
    layers.md
    locking.md
    migration.md
    talks.md
    why_kernels.md
examples/
  basic.py
  kernels/
    cpp20-symbols/
      build.toml
      cpu/
        cpu.cpp
      flake.nix
      tests/
        __init__.py
        test_cpp20_symbols.py
      torch-ext/
        cpp20_symbols/
          __init__.py
        torch_binding.cpp
        torch_binding.h
    cutlass-gemm/
    cutlass-gemm-tvm-ffi/
      build.toml
      CARD.md
      flake.nix
      gemm_sycl.cpp
      gemm.cu
      tests/
        __init__.py
        conftest.py
        test_gemm.py
      tvm-ffi-ext/
        cutlass_gemm_tvm_ffi/
          __init__.py
        tvm_ffi_binding.cpp
      util.hh
      build.toml
      CARD.md
      flake.nix
      gemm_sycl.cpp
      gemm.cu
      tests/
        __init__.py
        test_gemm.py
      torch-ext/
        cutlass_gemm/
          __init__.py
        registration.h
        torch_binding.cpp
        torch_binding.h
    extra-data/
      build.toml
      CARD.md
      flake.nix
      relu_cpu/
        relu_cpu.cpp
      relu_cuda/
        relu.cu
      relu_metal/
        common.h
        relu.metal
        relu.mm
      relu_xpu/
        relu.cpp
      tests/
        __init__.py
        test_relu.py
      torch-ext/
        extra_data/
          __init__.py
          data.json
          layers/
            __init__.py
        torch_binding.cpp
        torch_binding.h
    flake.nix
    relu/
    relu-archs-subset/
      build.toml
      CARD.md
      flake.nix
      relu_cuda/
        relu.cu
      torch-ext/
        relu_archs_subset/
          __init__.py
        torch_binding.cpp
        torch_binding.h
    relu-backprop-compile/
      build.toml
      CARD.md
      flake.nix
      relu_cuda/
        relu.cu
      tests/
        __init__.py
        test_relu.py
      torch-ext/
        relu_backprop_compile/
          __init__.py
        torch_binding.cpp
        torch_binding.h
    relu-compiler-flags/
      build.toml
      CARD.md
      flake.nix
      relu_cuda/
        relu.cu
      relu_xpu/
        relu.cpp
      torch-ext/
        relu_compiler_flags/
          __init__.py
        torch_binding.cpp
        torch_binding.h
    relu-invalid-capability/
      build.toml
      CARD.md
      flake.nix
      relu_cuda/
        relu.cu
      torch-ext/
        relu_invalid_capability/
          __init__.py
        torch_binding.cpp
        torch_binding.h
    relu-metal-cpp/
      build.toml
      CARD.md
      flake.nix
      relu/
        common.h
        metallib_loader.mm
        relu_cpp.metal
        relu.cpp
      tests/
        __init__.py
        test_relu.py
      torch-ext/
        relu/
          __init__.py
        torch_binding.cpp
        torch_binding.h
    relu-nki/
      build.toml
      CARD.md
      torch-ext/
        relu_nki/
          __init__.py
          layers/
            __init__.py
    relu-specific-torch/
      build.toml
      CARD.md
      flake.nix
      relu_cuda/
        relu.cu
      tests/
        __init__.py
        test_relu.py
      torch-ext/
        relu_specific_torch/
          __init__.py
        torch_binding.cpp
        torch_binding.h
    relu-torch-bounds/
      build.toml
      CARD.md
      flake.nix
      relu_cuda/
        relu.cu
      tests/
        __init__.py
        test_relu.py
      torch-ext/
        relu/
          __init__.py
        torch_binding.cpp
        torch_binding.h
    relu-torch-stable-abi/
      build.toml
      CARD.md
      flake.nix
      relu_cpu/
        relu_cpu.cpp
      relu_cuda/
        relu.cu
      relu_metal/
        common.h
        relu.metal
        relu.mm
      relu_xpu/
        relu.cpp
      tests/
        __init__.py
        conftest.py
        test_relu.py
      torch-ext/
        relu_torch_stable_abi/
          __init__.py
          layers/
            __init__.py
        torch_binding.cpp
        torch_binding.h
    relu-tpu/
      build.toml
      CARD.md
      flake.nix
      torch-ext/
        relu_tpu/
          __init__.py
          layers/
            __init__.py
    relu-triton/
      build.toml
      CARD.md
      flake.nix
      tests/
        __init__.py
        conftest.py
        test_relu_triton.py
      torch-ext/
        relu_triton/
          __init__.py
          layers.py
          op.py
    relu-tvm-ffi/
    relu-tvm-ffi-compiler-flags/
      build.toml
      CARD.md
      flake.nix
      relu_cpu/
        relu_cpu.cpp
      relu_cuda/
        relu.cu
      relu_xpu/
        relu.cpp
      tvm-ffi-ext/
        relu_tvm_ffi_compiler_flags/
          __init__.py
          layers/
            __init__.py
        tvm_ffi_binding.cpp
      util.hh
      build.toml
      CARD.md
      flake.nix
      relu_cpu/
        relu_cpu.cpp
      relu_cuda/
        relu.cu
      relu_xpu/
        relu.cpp
      tests/
        __init__.py
        conftest.py
        test_relu.py
      tvm-ffi-ext/
        relu_tvm_ffi/
          __init__.py
          layers/
            __init__.py
        tvm_ffi_binding.cpp
      util.hh
      build.toml
      CARD.md
      flake.nix
      relu_cpu/
        relu_cpu.cpp
      relu_cuda/
        relu.cu
      relu_metal/
        common.h
        relu.metal
        relu.mm
      relu_xpu/
        relu.cpp
      tests/
        __init__.py
        conftest.py
        test_relu.py
      torch-ext/
        relu/
          __init__.py
          layers/
            __init__.py
        torch_binding.cpp
        torch_binding.h
    silu-and-mul/
    silu-and-mul-bad-registration/
      build.toml
      CARD.md
      flake.nix
      torch-ext/
        silu_and_mul_bad_registration/
          __init__.py
          layers.py
          op.py
      build.toml
      CARD.md
      flake.nix
      tests/
        test_silu_and_mul.py
      torch-ext/
        silu_and_mul/
          __init__.py
          layers.py
          op.py
    symbol-conflicts/
      build.toml
      conflicts_cpu/
        conflicts_cpu.cpp
      flake.nix
      torch-ext/
        symbol_conflicts/
          __init__.py
        torch_binding.cpp
        torch_binding.h
    symbol-conflicts2/
      build.toml
      conflicts_cpu/
        conflicts_cpu.cpp
      flake.nix
      tests/
... 575 more

More Flows by @huggingface.

VerifiedPlugin

Development

huggingface-skills

Hugging Face Skills are definitions for AI/ML tasks like dataset creation, model training, and evaluation.

26 skills

@huggingface

View Plugin

FAQ

kernels is a Claude Code plugin with 5 hand-picked skills for development work, indexed on Flowy. Install it with the command on its page. It includes cpu-kernels, cuda-kernels, rocm-kernels. Its skills do not fire on their own yet. Request auto-invocation to have Flowy route them as you prompt. Free and open source.

What is kernels?: The Kernel Hub allows Python libraries and applications to load compute kernels directly from the Hub.
How do I install kernels?: Run these in Claude Code: npx -y skills add huggingface/kernels --agent claude-code. Then prompt normally.
Does kernels auto-invoke its skills?: Not yet. It is indexed on Flowy as a plain plugin. Request auto-invocation on its page and Flowy will route its skills for you as you prompt.
Is kernels free?: Yes. Flowy is free and open source, with nothing gated. You can read every skill in full before you install.

kernels

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.

🚀 Quick Start

Install the kernels Python package with pip (requires torch>=2.5 and CUDA):

pip install kernels

Here is how you would use the activation kernels from the Hugging Face Hub:

import torch from kernels import get_kernel # Download optimized kernels from the Hugging Face hub activation = get_kernel("kernels-community/activation", version=1) # Random tensor x = torch.randn((10, 10), dtype=torch.float16, device="cuda") # Run the kernel y = torch.empty_like(x) activation.gelu_fast(y, x) print(y)

.github/ dependabot.yml ISSUE_TEMPLATE/ bug_report.yml feature_request.yml pull_request_template.md VOUCHED.td workflows/ build_documentation.yaml build_kernel_cpu.yaml build_kernel_macos.yaml build_kernel_rocm.yaml build_kernel_windows.yaml build_kernel_xpu.yaml build_kernel.yaml build_pr_documentation.yaml check_variants.yaml coverage_comment.yaml kernel-builder-cli-docs.yaml kernels_data_python_release.yaml lint.yml nix_checks.yml publish_kernels.yml rust.yaml security-audit.yml test_e2e.yaml test_extra_commands.yaml test_kernels.yaml test_python.yaml update_abi_symbols.yaml update_cache.yaml upload_pr_documentation.yaml vouch-check-pr.yaml .gitignore .pre-commit-config.yaml Cargo.lock Cargo.toml CONTRIBUTING.md default.nix docs/ source/ _toctree.yml api/ kernels.md layers.md basic-usage.md blog.md builder/ builder-cli.md agents-guide.md build-variants.md build.md design-nix-builder.md design-overview.md github-actions.md ide-setup.md local-dev.md metal.md security.md why-nix.md writing-kernels.md cli-benchmark.md cli-download.md cli-info.md cli-lock.md cli-skills.md cli-verify-signature.md cli-versions.md cli.md env.md faq.md index.md installation.md integrating-kernels.md kernel-requirements.md layers.md locking.md migration.md talks.md why_kernels.md examples/ basic.py kernels/ cpp20-symbols/ build.toml cpu/ cpu.cpp flake.nix tests/ __init__.py test_cpp20_symbols.py torch-ext/ cpp20_symbols/ __init__.py torch_binding.cpp torch_binding.h cutlass-gemm/ cutlass-gemm-tvm-ffi/ build.toml CARD.md flake.nix gemm_sycl.cpp gemm.cu tests/ __init__.py conftest.py test_gemm.py tvm-ffi-ext/ cutlass_gemm_tvm_ffi/ __init__.py tvm_ffi_binding.cpp util.hh build.toml CARD.md flake.nix gemm_sycl.cpp gemm.cu tests/ __init__.py test_gemm.py torch-ext/ cutlass_gemm/ __init__.py registration.h torch_binding.cpp torch_binding.h extra-data/ build.toml CARD.md flake.nix relu_cpu/ relu_cpu.cpp relu_cuda/ relu.cu relu_metal/ common.h relu.metal relu.mm relu_xpu/ relu.cpp tests/ __init__.py test_relu.py torch-ext/ extra_data/ __init__.py data.json layers/ __init__.py torch_binding.cpp torch_binding.h flake.nix relu/ relu-archs-subset/ build.toml CARD.md flake.nix relu_cuda/ relu.cu torch-ext/ relu_archs_subset/ __init__.py torch_binding.cpp torch_binding.h relu-backprop-compile/ build.toml CARD.md flake.nix relu_cuda/ relu.cu tests/ __init__.py test_relu.py torch-ext/ relu_backprop_compile/ __init__.py torch_binding.cpp torch_binding.h relu-compiler-flags/ build.toml CARD.md flake.nix relu_cuda/ relu.cu relu_xpu/ relu.cpp torch-ext/ relu_compiler_flags/ __init__.py torch_binding.cpp torch_binding.h relu-invalid-capability/ build.toml CARD.md flake.nix relu_cuda/ relu.cu torch-ext/ relu_invalid_capability/ __init__.py torch_binding.cpp torch_binding.h relu-metal-cpp/ build.toml CARD.md flake.nix relu/ common.h metallib_loader.mm relu_cpp.metal relu.cpp tests/ __init__.py test_relu.py torch-ext/ relu/ __init__.py torch_binding.cpp torch_binding.h relu-nki/ build.toml CARD.md torch-ext/ relu_nki/ __init__.py layers/ __init__.py relu-specific-torch/ build.toml CARD.md flake.nix relu_cuda/ relu.cu tests/ __init__.py test_relu.py torch-ext/ relu_specific_torch/ __init__.py torch_binding.cpp torch_binding.h relu-torch-bounds/ build.toml CARD.md flake.nix relu_cuda/ relu.cu tests/ __init__.py test_relu.py torch-ext/ relu/ __init__.py torch_binding.cpp torch_binding.h relu-torch-stable-abi/ build.toml CARD.md flake.nix relu_cpu/ relu_cpu.cpp relu_cuda/ relu.cu relu_metal/ common.h relu.metal relu.mm relu_xpu/ relu.cpp tests/ __init__.py conftest.py test_relu.py torch-ext/ relu_torch_stable_abi/ __init__.py layers/ __init__.py torch_binding.cpp torch_binding.h relu-tpu/ build.toml CARD.md flake.nix torch-ext/ relu_tpu/ __init__.py layers/ __init__.py relu-triton/ build.toml CARD.md flake.nix tests/ __init__.py conftest.py test_relu_triton.py torch-ext/ relu_triton/ __init__.py layers.py op.py relu-tvm-ffi/ relu-tvm-ffi-compiler-flags/ build.toml CARD.md flake.nix relu_cpu/ relu_cpu.cpp relu_cuda/ relu.cu relu_xpu/ relu.cpp tvm-ffi-ext/ relu_tvm_ffi_compiler_flags/ __init__.py layers/ __init__.py tvm_ffi_binding.cpp util.hh build.toml CARD.md flake.nix relu_cpu/ relu_cpu.cpp relu_cuda/ relu.cu relu_xpu/ relu.cpp tests/ __init__.py conftest.py test_relu.py tvm-ffi-ext/ relu_tvm_ffi/ __init__.py layers/ __init__.py tvm_ffi_binding.cpp util.hh build.toml CARD.md flake.nix relu_cpu/ relu_cpu.cpp relu_cuda/ relu.cu relu_metal/ common.h relu.metal relu.mm relu_xpu/ relu.cpp tests/ __init__.py conftest.py test_relu.py torch-ext/ relu/ __init__.py layers/ __init__.py torch_binding.cpp torch_binding.h silu-and-mul/ silu-and-mul-bad-registration/ build.toml CARD.md flake.nix torch-ext/ silu_and_mul_bad_registration/ __init__.py layers.py op.py build.toml CARD.md flake.nix tests/ test_silu_and_mul.py torch-ext/ silu_and_mul/ __init__.py layers.py op.py symbol-conflicts/ build.toml conflicts_cpu/ conflicts_cpu.cpp flake.nix torch-ext/ symbol_conflicts/ __init__.py torch_binding.cpp torch_binding.h symbol-conflicts2/ build.toml conflicts_cpu/ conflicts_cpu.cpp flake.nix tests/ ... 575 more