skills.rest
sqjian
sqjianCOMMUNITY
Explore all skills by this author →
GitHub

sglang-torch-profiler-analysis

Analyze SGLang torch-profiler traces for kernel overlaps and fusion opportunities.

Updated May 14, 2026
Search Tags:#analysis#kernel#triage#overlap#profiler#sglang#fusion
One-click install
npx skills add https://github.com/sqjian/sglang --skill sglang-torch-profiler-analysis-sqjian
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: sglang-torch-profiler-analysis
Source: https://github.com/sqjian/sglang/tree/main/.claude/skills/sglang-torch-profiler-analysis
Command: npx skills add https://github.com/sqjian/sglang --skill sglang-torch-profiler-analysis-sqjian

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires triage_kernel_helpers, triage_overlap_helpers, profile_common, and includes scripts (resource) and references (resource) components.

What problem does it solve?

Identify and triage kernel-level overlaps and fusion opportunities in SGLang torch-profiler traces.

Core Features & Use Cases

  • Kernel-level tables including kernel time, stage, and location for quick debugging.
  • Overlap opportunity analysis that highlights cross-stream and dependency risks.
  • Fusion pattern assessment surfacing existing, PR-backed, and potential fusions with guidance.

Quick Start

Run the triage against a local trace directory or a running SGLang server to generate the three tables.

Frequently Asked Questions about sglang-torch-profiler-analysis

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

FAQPage Schema
How do I identify kernel-level overlap opportunities in SGLang torch-profiler traces?▼

To identify kernel-level overlap opportunities in SGLang torch-profiler traces, this Skill analyzes cross-stream and dependency risks, rendering a compact overlap-opportunity table for quick diagnosis.

What is the best way to find kernel fusion patterns in SGLang torch profiler outputs?▼

Finding kernel fusion patterns in SGLang torch profiler outputs involves assessing existing, PR-backed, and potential fusions, which this Skill surfaces in a fuse pattern table with guidance.

Can I triage kernel performance issues using a local trace directory or a live SGLang server?▼

Yes, you can triage kernel performance issues by loading trace data from either local trace directories or a running SGLang server to generate a deterministic triage report.

Does this Skill support dual-trace mapping for SGLang torch-profiler analysis?▼

Yes, this Skill supports dual-trace mapping and formal pairs for SGLang torch-profiler analysis, alongside single-trace inputs, to render kernel, overlap-opportunity, and fuse pattern tables.

How do I debug kernel time and stage location data from SGLang torch-profiler traces?▼

To debug kernel time and stage location data from SGLang torch-profiler traces, this Skill generates a kernel-level table that includes kernel time, stage, and location for quick debugging.

What are the limitations of analyzing SGLang torch-profiler traces for fusion opportunities?▼

The analysis of SGLang torch-profiler traces for fusion opportunities is limited to rendering deterministic reports based on kernel, overlap, and fuse pattern tables, focusing on quick diagnosis rather than automated code optimization.

Related Skills

Explore More Software Engineering Skills→
BBufBBuf
721

llm-torch-profiler-analysis

Analyze LLM torch-profiler traces to produce kernel, overlap, and fuse tables.

Community
Intermediate
BBufBBuf
721

sglang-sota-performance

Benchmark models across SGLang, vLLM, and TensorRT-LLM to identify performance gaps.

Community
Advanced
annealing-inversionannealing-inversion

llm-torch-profiler-analysis

Analyze LLM torch-profiler traces into kernel, overlap, and fuse pattern reports.

Community
Advanced
OpenMOSSOpenMOSS
410

sglang-torch-profiler-analysis

Triage SGLang torch-profiler traces into kernel, overlap, and fuse-pattern tables.

Official
Advanced