triton-cuda-attention

Community

Triton-CUDA attention with Flash Attention variants

Authorxchang1121
Version1.0.0
Installs0

System Documentation

What problem does it solve?

Efficiently implement and validate Triton-CUDA attention mechanisms, including Flash Attention variants, online Softmax, and robust kernel patterns.

Core Features & Use Cases

  • Proven Flash Attention integration: complete Triton-CUDA implementation examples with variations (Causal, GQA, MQA, RoPE) and practical usage notes.
  • Online Softmax: memory-efficient, numerically stable softmax for block-wise attention.
  • Common pitfalls & debugging: guidance on typical mistakes and how to verify correctness across kernels and host code.

Quick Start

Build and run the Triton-CUDA attention example to verify Flash Attention variants on a sample input.

Dependency Matrix

Required Modules

None required

Components

Standard package

💻 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: triton-cuda-attention
Download link: https://github.com/xchang1121/AutoResearch-CC-hook/archive/main.zip#triton-cuda-attention

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 471,000+ vetted skills library on demand.