BBuf/sglang-hunyuan3-preview-optimization
skills/model-optimization/sglang/sglang-hunyuan3-preview-optimization/SKILL.md
PR-backed and current-main optimization manual for Tencent Hunyuan 3 Preview in SGLang. Use when an engineer needs to audit or extend Hunyuan3 Preview cookbook recipes, BF16 MoE hardware sizing, H200/B200/B300/GB300 command generation, MTP/EAGLE flags, `hunyuan` reasoning/tool parsers, Blackwell attention backend selection, or trust-remote-code launch guidance.
$ install --global
skillsauthnpx skillsauth add BBuf/AI-Infra-Auto-Driven-SKILLS sglang-hunyuan3-preview-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, 2:35 AM213.7s2 files scanned
SKILL.md
- name:
- sglang-hunyuan3-preview-optimization
- description:
- PR-backed and current-main optimization manual for Tencent Hunyuan 3 Preview in SGLang. Use when an engineer needs to audit or extend Hunyuan3 Preview cookbook recipes, BF16 MoE hardware sizing, H200/B200/B300/GB300 command generation, MTP/EAGLE flags, `hunyuan` reasoning/tool parsers, Blackwell attention backend selection, or trust-remote-code launch guidance.
SGLang Hunyuan 3 Preview Optimization
Overview
Hunyuan 3 Preview currently enters SGLang through docs/cookbook support with a copyable command generator. The PR is docs-only, but the generator records critical serving assumptions: BF16 weights, TP sizing by GPU memory, parser flags, MTP/EAGLE toggles, Blackwell trtllm_mha, and --trust-remote-code.
Current evidence snapshot:
- SGLang
origin/main:bca3dd958on2026-04-24 - Primary PR: #23532
- Docs:
docs_new/cookbook/autoregressive/Tencent/Hunyuan3-Preview.mdx - Command generator:
docs_new/src/snippets/autoregressive/hunyuan3-preview-deployment.jsx
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.
Before You Change Anything
Capture:
- hardware: H200, B200, B300, or GB300
- TP and memory fraction
- whether MTP is enabled and
SGLANG_ENABLE_SPEC_V2=1 - parser flags:
--reasoning-parser hunyuan,--tool-call-parser hunyuan - whether Blackwell requires
--attention-backend trtllm_mha - whether
--trust-remote-codeis present
Core Principle
Treat the command generator as deployment data.
- H200/B200 use TP=8 for BF16.
- B300/GB300 use TP=4 for BF16.
- A100/H100 80GB single-node is not supported by the documented recipe.
- Blackwell must pass
--attention-backend trtllm_mha.
PR Dossier Rule
Before adding Hunyuan3 Preview evidence, open the PR diff/source and update references/pr-history.md with motivation, implementation, code/config excerpts, reviewed files, and validation implications.
Validation Lanes
- H200 and B200 command generation with TP=8.
- B300 and GB300 command generation with TP=4 and
trtllm_mha. - reasoning/tool parser toggles.
- MTP toggle adding
SGLANG_ENABLE_SPEC_V2=1and EAGLE flags. --trust-remote-codealways present.
References
references/pr-history.md: diff-reviewed Hunyuan3 Preview PR cards.
Related Skills
BBuf/sglang-humanize-review
development
VerifiedTrustedCommunity
Perform SGLang code review in the style of human maintainers by consulting the full non-agent PR review episode corpus from project start through the latest refresh (June 2026), including inline review threads, top-level PR comments, review submissions, original multilingual text, and multi-round discussions. Use when reviewing SGLang PRs, diffs, patches, or local changes for correctness, tests, performance, GPU/runtime risks, API compatibility, and maintainability.
531SKILL.mdUpdated May 21, 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.
531SKILL.mdUpdated May 17, 2026
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