anthony-maio

Triton Quantized Block Scaled Gemm — AI skill for Claude Code

AI community

Teach an AI agent how to implement block-scaled (microscaling) quantized matmul kernels in Triton.

How to install Triton Quantized Block Scaled Gemm

Installs to ~/.claude/skills/anthony-maio-triton-skills-triton-quantized-block-scaled-gemm/SKILL.md

Terminal

mkdir -p ~/.claude/skills/anthony-maio-triton-skills-triton-quantized-block-scaled-gemm && curl -fsSL https://raw.githubusercontent.com/anthony-maio/triton-skills/HEAD/skills/triton-quantized-block-scaled-gemm.md -o ~/.claude/skills/anthony-maio-triton-skills-triton-quantized-block-scaled-gemm/SKILL.md

Restart Claude Code, or start a new session, for it to be picked up.

What Triton Quantized Block Scaled Gemm does

name: triton-quantized-block-scaled-gemm description: Teach an AI agent how to implement block-scaled (microscaling) quantized matmul kernels in Triton.

Quantized & Block-Scaled Matmul Kernels in Triton

**Targets:** Triton >= 3.0; `tl.dot_scaled` requires SM100+/CDNA4; dequantize fallback works on SM70+/CDNA2+

Overview This guide explains how to implement low-precision block-scaled matrix multiplication in Triton for mxfp4/mxfp8/nvfp4 formats. It covers scale tensor layouts (OCP m

Alternatives in AI

Gsap Skills — Official AI skills for GSAP 1.7k ★
Hol Guard — Open-source antivirus for AI agents: block risky tools, secret access, prompt injection, malicious packages, M 485 ★
Prismor — Self-hosted runtime control plane for AI agents 334 ★