skills.rest
BBuf
BBufCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

sglang-glm45-optimization

Optimize GLM-4.5 MoE loading and runtime tuning in SGLang.

721|65|Updated Apr 1, 2026
Search Tags:#validation#optimization#sglang#cookbook#moe#glm-4.5
One-click install
npx skills add https://github.com/BBuf/AI-Infra-Auto-Driven-SKILLS --skill sglang-glm45-optimization
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: sglang-glm45-optimization
Source: https://github.com/BBuf/AI-Infra-Auto-Driven-SKILLS/tree/main/skills/model-optimization/sglang/sglang-glm45-optimization
Command: npx skills add https://github.com/BBuf/AI-Infra-Auto-Driven-SKILLS --skill sglang-glm45-optimization

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

PR-backed and current-main optimization manual for GLM-4.5 MoE in SGLang. Use when Codex needs to recover, extend, or audit GLM-4.5 MoE loading, A2A/DeepEP, reduce-scatter behavior, NVFP4 padding, tool parser behavior, AMD/NPU/Blackwell validation, or GLM-4.5 cookbook recipes.

Core Features & Use Cases

  • PR-backed optimization guidance for GLM-4.5 MoE loading, A2A/DeepEP, and runtime tuning.
  • Evidence-driven change rules: read PR diffs and dossier history to implement safe improvements.
  • Validation-focused playbook: cross-version checks, backport considerations, and deployment cookbook references.

Quick Start

Run the GLM-4.5 MoE optimization workflow on a GLM-4.5 model in SGLang.

Frequently Asked Questions about sglang-glm45-optimization

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I optimize GLM-4.5 MoE loading in SGLang?▼

To optimize GLM-4.5 MoE loading in SGLang, use PR-backed guidance to audit loading mechanisms, implement evidence-driven change rules from PR diffs, and apply cookbook deployment recipes for safe runtime improvements.

What is the best way to tune A2A and DeepEP for GLM-4.5?▼

Tuning A2A and DeepEP for GLM-4.5 involves applying PR-backed optimization guidance to adjust reduce-scatter behavior, validating changes against dossier history to ensure non-destructive improvements across runtime versions.

How does SGLang handle NVFP4 padding for GLM-4.5 models?▼

SGLang handles NVFP4 padding for GLM-4.5 models by applying evidence-based change rules derived from PR diffs, ensuring safe improvements and cross-version runtime validation for both GLM-4.5 and GLM-4.5-Air.

Can I use this SGLang optimization playbook for GLM-4.5-Air validation?▼

Yes, this SGLang optimization playbook supports GLM-4.5-Air validation by providing cross-version runtime checks, backport considerations, and specific deployment cookbook recipes for comprehensive model validation.

Why do I need to adjust tool parser behavior in SGLang for GLM-4.5?▼

Adjusting tool parser behavior in SGLang for GLM-4.5 ensures compatibility and safe runtime execution, utilizing PR-backed optimization guidance to maintain non-destructive changes during MoE deployment and validation.

What are the limitations when optimizing GLM-4.5 on AMD or NPU environments?▼

Optimizing GLM-4.5 on AMD, NPU, or Blackwell environments requires cross-version runtime validation and PR-dossier reviews to navigate hardware-specific limitations and ensure evidence-based, non-destructive change rules.

Related Skills

Explore More Software Engineering Skills→
BBufBBuf
721

vllm-glm45-optimization

Audit PR diffs and generate release-ready dossiers for GLM-4.5 series in vLLM.

Community
Advanced
BBufBBuf
721

vllm-glm46-glm47-optimization

Map GLM-4.6/4.7 optimization PRs to vLLM runtime files and tooling.

Community
Advanced
BBufBBuf
721

sglang-deepseek-v31-optimization

Optimize DeepSeek V3.1 tool calling, thinking mode, chat templates, and streaming parser behavior in SGLang.

Community
Advanced
BBufBBuf
721

sglang-glm5-glm51-optimization

Identify and vet GLM-5/5.1 optimization changes in SGLang PR diffs.

Community
Advanced