sglang-mixtral-quark-int4fp8-moe-optimization

Community

Optimize Mixtral INT4/FP8 MoE for AMD hardware.

AuthorBBuf
Version1.0.0
Installs0

System Documentation

What problem does it solve?

This Skill provides a production-ready workflow for online INT4-FP8 MoE quantization of Mixtral MoE models on AMD ROCm hardware, enabling efficient deployment and testing.

Core Features & Use Cases

  • Online quantization path for Mixtral-8x7B MoE checkpoints on ROCm/AMD.
  • Quantizes high-precision MoE weights to packed INT4 and stores per-weight scales for FP8-style MoE execution.
  • Guides auditing and extension of Mixtral AMD quantization in PR workflows, including GSM8K regression tests and diff history.

Quick Start

Start by configuring the AMD online int4fp8_moe quantization workflow for Mixtral-8x7B and run the associated GSM8K regression test.

Dependency Matrix

Required Modules

None required

Components

references

💻 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: sglang-mixtral-quark-int4fp8-moe-optimization
Download link: https://github.com/BBuf/AI-Infra-Auto-Driven-SKILLS/archive/main.zip#sglang-mixtral-quark-int4fp8-moe-optimization

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.