BBuf/vllm-llama4-optimization
skills/model-optimization/vllm/vllm-llama4-optimization/SKILL.md
PR-backed optimization manual for Llama 4 in vLLM. Use when an engineer needs to audit, debug, extend, or document Llama4 text and multimodal runtime, FP8/FP4 quantization, router behavior, long-context attention, and Eagle support.
$ install --global
skillsauthnpx skillsauth add BBuf/AI-Infra-Auto-Driven-SKILLS vllm-llama4-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 AM106.5s1 file scanned
SKILL.md
- name:
- vllm-llama4-optimization
- description:
- PR-backed optimization manual for Llama 4 in vLLM. Use when an engineer needs to audit, debug, extend, or document Llama4 text and multimodal runtime, FP8/FP4 quantization, router behavior, long-context attention, and Eagle support.
vLLM Llama 4 Optimization
Overview
This skill covers Llama4 text and multimodal runtime, FP8/FP4 quantization, router behavior, long-context attention, and Eagle support.
Evidence snapshot:
- vLLM mainline checked around
0f7be0f2f76814f80f9091220a5fbbb53912ad00 - Canonical PR notes:
references/pr-history.md - History mirrors:
model-pr-optimization-history/vllm/llama4/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/llama4.pyvllm/vllm/model_executor/models/mllama4.pyvllm/vllm/model_executor/models/llama4_eagle.py
Current Main Summary
- Llama4 is mature on the vLLM side but still sensitive to quantized MoE and long-context backend selection.
- The multimodal path adds a separate vision-rotary and processor validation surface.
Key Landed PRs
- #16104
Support Llama4 in vLLM: Initial Llama4 landing. - #20419
Enable ModelOpt Llama4 fp8 checkpoint deployment: Added ModelOpt FP8 coverage. - #20591
Llama4 EAGLE Support: Opened speculative decoding for Llama4. - #22511
Fix Llama4 FlashInfer FP4 MoE issues: Stabilized the FP4 MoE path. - #25889
Fix misplaced dtype cast in Llama4VisionRotaryEmbedding: Patched a multimodal rotary bug.
Validation Lanes
- Startup on the current vLLM mainline checkpoint.
- Re-run the parser, quantization, MTP, or multimodal lane that matches this family.
- Re-check tests after touching loader or processor code.
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