BBuf/vllm-qwen3-core-optimization
skills/model-optimization/vllm/vllm-qwen3-core-optimization/SKILL.md
PR-backed optimization manual for Qwen3 Core in vLLM. Use when an engineer needs to audit, debug, extend, or document Qwen3 dense, Qwen3 MoE, embeddings/rerankers, GGUF/GPTQ/ModelOpt quant paths, and Eagle3 speculative decoding in vLLM.
$ install --global
skillsauthnpx skillsauth add BBuf/AI-Infra-Auto-Driven-SKILLS vllm-qwen3-core-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, 2:44 AM257.6s2 files scanned
SKILL.md
- name:
- vllm-qwen3-core-optimization
- description:
- PR-backed optimization manual for Qwen3 Core in vLLM. Use when an engineer needs to audit, debug, extend, or document Qwen3 dense, Qwen3 MoE, embeddings/rerankers, GGUF/GPTQ/ModelOpt quant paths, and Eagle3 speculative decoding in vLLM.
vLLM Qwen3 Core Optimization
Overview
This skill covers Qwen3 dense, Qwen3 MoE, embeddings/rerankers, GGUF/GPTQ/ModelOpt quant paths, and Eagle3 speculative decoding 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/qwen3-core/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/qwen3.pyvllm/vllm/model_executor/models/qwen3_moe.pyvllm/vllm/model_executor/models/voyage.py
Current Main Summary
- Qwen3 and Qwen3MoE are first-class mainline runtimes in vLLM.
- The highest-risk regressions show up in quantized MoE weight loading, packed-module mappings, and embedding / reranker special cases.
Key Landed PRs
- #15289
Add Qwen3 and Qwen3MoE: Initial Qwen3 dense and MoE support landed here. - #19260
Support Qwen3 Embedding & Reranker: Extended the family to bidirectional embedding / reranker models. - #19598
Skip loading extra parameters for modelopt Qwen3 MoE model: Fixed a concrete ModelOpt launch failure on Qwen3 MoE. - #22017
KeyError for Qwen3-MoE with GPTQ on ROCm: Closed a GPTQ loading failure in the Qwen3 MoE path. - #22785
Fix GGUF loader for Qwen3 MoE: Made the Qwen3 MoE loader accept GGUF weights again. - #23490
Fix Qwen3 MoE GPTQ inference: Patched runtime correctness after GPTQ startup succeeded. - #26485
Add EAGLE-3 Speculative Decoding Support for Qwen3 MoE: Enabled the draft-model path on top of the base Qwen3 MoE runtime.
Open Optimization Items
- Re-run PR search before claiming new support beyond the checked mainline commit.
Validation Lanes
- Dense and MoE BF16 launch.
- Embedding / reranker special-case loading.
- GPTQ / GGUF / ModelOpt startup on Qwen3 MoE.
- EAGLE-3 acceptance-rate sanity on supported checkpoints.
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