h100-diffusers-kernels

Community

Boost CUDA kernels for Diffusers on H100

Authorburtenshaw
Version1.0.0
Installs0

System Documentation

What problem does it solve?

This Skill provides patterns and guidance for writing optimized CUDA kernels on NVIDIA H100 GPUs (sm_90) to maximize performance of diffusion models using the HuggingFace diffusers library.

Core Features & Use Cases

  • Kernel templates and patterns for RMSNorm, RoPE (1D/3D), GEGLU, SwiGLU, and AdaLN tailored to H100.
  • Diffusers integration guidance for LTX-Video, Stable Diffusion, and DiT, including PyTorch bindings and example workflows.
  • Performance optimization guidance covering block sizes, shared memory usage, memory bandwidth considerations, and profiling commands.

Quick Start

Install the minimal working patch example, load the LTX-Video pipeline on CUDA, and run the kernel injection flow to apply the optimized kernels.

Dependency Matrix

Required Modules

torchdiffusersltx_kernels

Components

scriptsreferences

💻 Claude Code Installation

Recommended: Let Claude install automatically. Simply copy and paste the text below to Claude Code.

Please help me install this Skill:
Name: h100-diffusers-kernels
Download link: https://github.com/burtenshaw/kernel-skill/archive/main.zip#h100-diffusers-kernels

Please download this .zip file, extract it, and install it in the .claude/skills/ directory.
View Source Repository

Agent Skills Search Helper

Install a tiny helper to your Agent, search and equip skill from 223,000+ vetted skills library on demand.