skills.rest
dyyoungg
dyyounggCOMMUNITY
Explore all skills by this author →
GitHub

llm-torch-profiler-analysis

Analyze LLM torch-profiler traces to identify kernel, overlap, and fusion patterns.

1|Updated May 8, 2026
Search Tags:#llm#profiling#triage#kernel-analysis#torch-profiler#fusion-patterns#overlap-patterns
One-click install
npx skills add https://github.com/dyyoungg/sglang-dev --skill llm-torch-profiler-analysis-dyyoungg
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: llm-torch-profiler-analysis
Source: https://github.com/dyyoungg/sglang-dev/tree/main/.claude/skills/llm-torch-profiler-analysis
Command: npx skills add https://github.com/dyyoungg/sglang-dev --skill llm-torch-profiler-analysis-dyyoungg

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

This Skill enables data-driven triage of LLM torch-profiler traces across SGLang, vLLM, and TensorRT-LLM, surfacing actionable kernel bottlenecks, overlap opportunities, and fusion patterns to accelerate optimization work.

Core Features & Use Cases

  • Triages both single-trace and two-trace mapping/formal analysis to generate kernel, overlap-opportunity, and fusion patterns.
  • Produces three tables and a concise summary suitable for documentation and handoff to engineering teams.
  • Enables cross-backend comparison to guide performance tuning and backend-appropriate fusion decisions.

Quick Start

Run analyze_llm_torch_profile.py on a trace directory to produce the triage report.

Frequently Asked Questions about llm-torch-profiler-analysis

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

FAQPage Schema
How do I triage torch-profiler traces to find kernel bottlenecks in vLLM?▼

To triage torch-profiler traces for vLLM, you analyze trace data to identify kernel bottlenecks, overlap opportunities, and fusion patterns. This Skill processes single-trace data to generate triage tables and a summary.

What is the best way to compare profiling results between SGLang and TensorRT-LLM?▼

Comparing profiling results between SGLang and TensorRT-LLM is done through two-trace mapping and formal analysis. This reveals cross-backend kernel, overlap, and fusion patterns to guide performance tuning and fusion decisions.

Does this triage tool support both single-trace and two-trace analyses?▼

Yes, this triage tool supports both single-trace analysis and two-trace mapping. It renders three tables covering kernel, overlap, and fusion patterns, plus a summary suitable for documentation and engineering handoff.

How do I start analyzing an LLM torch-profiler trace directory?▼

To start analyzing an LLM torch-profiler trace, run the analyze_llm_torch_profile.py script on your trace directory. This produces a triage report containing kernel, overlap, and fusion pattern tables.

Can I identify fusion patterns and overlap opportunities from torch.profiler output?▼

Yes, you can identify fusion patterns and overlap opportunities from torch.profiler output. The Skill analyzes traces to surface these actionable patterns alongside kernel bottlenecks to accelerate optimization work.

Related Skills

Explore More Software Engineering Skills→
hey-konghey-kong

llm-torch-profiler-analysis

Analyze torch.profiler traces to generate kernel, overlap, and fuse-pattern tables.

Community
Advanced
sqjiansqjian

sglang-torch-profiler-analysis

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

Community
Intermediate
QingHuan-6QingHuan-6

llm-serving-auto-benchmark

Benchmark SGLang, vLLM, and TensorRT-LLM serving to produce Markdown/CSV summaries.

Community
Advanced
OpenMOSSOpenMOSS
410

sglang-torch-profiler-analysis

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

Official
Advanced