BBuf/vllm-glm45-optimization
skills/model-optimization/vllm/vllm-glm45-optimization/SKILL.md
PR-backed optimization manual for GLM-4.5 / 4.5V in vLLM. Use when an engineer needs to audit, debug, extend, or document GLM-4.5 text, GLM-4.5V, GLM-4.5-Air, shared MoE routing, and tool/reasoning parser behavior in vLLM.
$ install --global
skillsauthnpx skillsauth add BBuf/AI-Infra-Auto-Driven-SKILLS vllm-glm45-optimizationInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: May 2, 2026, 2:39 AM140.2s2 files scanned
SKILL.md
- name:
- vllm-glm45-optimization
- description:
- PR-backed optimization manual for GLM-4.5 / 4.5V in vLLM. Use when an engineer needs to audit, debug, extend, or document GLM-4.5 text, GLM-4.5V, GLM-4.5-Air, shared MoE routing, and tool/reasoning parser behavior in vLLM.
vLLM GLM-4.5 / 4.5V Optimization
Overview
This skill covers GLM-4.5 text, GLM-4.5V, GLM-4.5-Air, shared MoE routing, and tool/reasoning parser behavior in vLLM.
Evidence snapshot:
- vLLM mainline checked around
0f7be0f2f76814f80f9091220a5fbbb53912ad00 - Support status: supported on current mainline
- Canonical PR notes:
references/pr-history.md - History mirrors:
model-pr-optimization-history/vllm/glm45/README.zh.mdandREADME.en.md
Non-Negotiable Evidence Rule
Use skills/model-optimization/model-pr-diff-dossier/SKILL.md as the production bar.
Every PR cited for this family must be based on diff reading, not only PR titles.
Runtime Surfaces
vllm/vllm/model_executor/models/glm4.pyvllm/vllm/model_executor/models/glm4_moe.pyvllm/vllm/model_executor/models/glm4v.py
Current Main Summary
- The GLM-4.5 lane is where vLLM reorganized the GLM family around text, MoE, and vision variants.
- Most regressions are in MoE gate behavior, tie-word-embedding policy, and vendor-specific fused MoE tuning.
Key Landed PRs
- #22171
Modify the organization of GLM series: Reworked the family layout so 4.5-era models reused a cleaner GLM structure. - #22460
not tie_word_embeddings for glm-4.5 and glm-4.5v: Aligned the loader with the real 4.5 checkpoint contract instead of forcing tied embeddings. - #22832
Modify the gate implementation of glm4_moe: Changed the GLM4.5 MoE gating path used by text and VL variants. - #23695
Add triton fused moe config for GLM-4.5-Air-FP8 on B200: Added a production kernel-tuning lane for the 4.5 Air FP8 deployment path. - #24589
Add documentation for GLM-4.5 series tool-calling and reasoning parser: Codified the parser choices needed to serve 4.5 reasoning / tool checkpoints correctly.
Open Optimization Items
- Re-run PR search before claiming new support beyond the checked mainline commit.
Validation Lanes
- GLM-4.5 and GLM-4.5-Air text generation.
- GLM-4.5V startup with image inputs.
- Tool calling and reasoning parser smoke tests.
- FP8 fused-MoE launch on Blackwell where relevant.
Related Skills
BBuf/vllm-sota-humanize-loop
development
VerifiedTrustedCommunity
Run an autonomous Humanize-governed vLLM SOTA performance loop for one LLM model: first perform the fixed fair vLLM/SGLang/TensorRT-LLM deployment search and benchmark, then start one RLCR loop that repeatedly decides the gap, profiles the current bottleneck, runs layer/kernel pipeline analysis, patches vLLM code, optionally uses ncu-report-skill for kernel evidence, and revalidates until vLLM matches or beats the best observed framework under the same workload and SLA.
423SKILL.mdUpdated May 27, 2026
BBuf/llm-pipeline-analysis
devops
VerifiedTrustedCommunity
Inspect LLM torch profiler traces at forward-pass, layer, and kernel level. Use when you need layer timings, anchor-kernel boundaries, representative kernel flows, or Perfetto time ranges.
423SKILL.mdUpdated May 21, 2026
BBuf/sglang-sota-humanize-loop
development
VerifiedTrustedCommunity
Run an autonomous Humanize-governed SGLang SOTA performance loop for one LLM model: first perform the fixed fair SGLang/vLLM/TensorRT-LLM deployment search and benchmark, then start one RLCR loop that repeatedly decides the gap, profiles the current bottleneck, runs layer/kernel pipeline analysis, patches SGLang code, optionally uses ncu-report-skill for kernel evidence, and revalidates until SGLang matches or beats the best observed framework under the same workload and SLA.
423SKILL.mdUpdated May 17, 2026
BBuf/model-pr-history-knowledge
documentation
VerifiedTrustedCommunity
Use when an SGLang, vLLM, or TensorRT-LLM serving/model optimization task needs prior model-family PR evidence. Query and read the PR-driven history docs under model-pr-optimization-history before choosing source paths, fast paths, kernel/fusion ideas, regression risks, or validation lanes.
423SKILL.mdUpdated May 17, 2026