skills.rest
annealing-inversion
annealing-inversionCOMMUNITY
Explore all skills by this author →
GitHub

llm-torch-profiler-analysis

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

Updated Jun 4, 2026
Search Tags:#llm#triage#overlap#torch-profiler#kernel-table#fuse-patterns
One-click install
npx skills add https://github.com/annealing-inversion/sglang-kimi-deferral --skill llm-torch-profiler-analysis-annealing-inversion
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: llm-torch-profiler-analysis
Source: https://github.com/annealing-inversion/sglang-kimi-deferral/tree/main/.claude/skills/llm-torch-profiler-analysis
Command: npx skills add https://github.com/annealing-inversion/sglang-kimi-deferral --skill llm-torch-profiler-analysis-annealing-inversion

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

This Skill consolidates and analyzes LLM torch-profiler traces to produce a cohesive three-table triage (kernel, overlap, and fuse pattern) across SGLang, vLLM, and TensorRT-LLM, enabling quicker diagnosis of performance bottlenecks.

Core Features & Use Cases

  • Unified, cross-framework triage: analyze single-trace and dual-trace profiling data to surface actionable kernel and fusion opportunities.
  • Live and offline analysis: supports profiling live servers or validating existing trace directories to produce a consistent three-table report.
  • Real-world use: a performance engineer can snapshot a running model server, run the triage, and receive a ready-to-embed report for optimization planning.

Quick Start

Run the analyze_llm_torch_profile.py script on a trace to generate the three triage tables.

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 LLM torch-profiler traces to find performance bottlenecks?▼

To triage LLM torch-profiler traces, you run the analyze_llm_torch_profile.py script to extract trace data and compute a unified three-table markdown summary covering kernel, overlap, and fuse pattern reports.

Can I analyze torch-profiler traces from different LLM frameworks using a single tool?▼

Yes, you can analyze torch-profiler traces from SGLang, vLLM, and TensorRT-LLM using a single tool that computes cross-framework triage tables to surface actionable kernel and fusion opportunities.

How do I profile a live running model server and generate a triage report?▼

You can profile a live running model server by executing the analyze_llm_torch_profile.py script in live mode, which snapshots the active server to produce a consistent three-table triage report.

Does this torch-profiler analysis support comparing two different trace files?▼

Yes, the torch-profiler analysis supports a two-trace workflow, allowing you to compare profiling data across different runs or configurations to identify performance regressions and optimization opportunities.

What format does the torch-profiler triage report output for optimization planning?▼

The torch-profiler triage report outputs a ready-to-embed three-table markdown summary, rendering kernel, overlap, and fuse pattern triage tables for optimization planning.

Related Skills

Explore More Software Engineering Skills→
dyyounggdyyoungg

llm-torch-profiler-analysis

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

Community
Advanced
smith-hornsmith-horn
75

performance-analysis

Detect and profile performance bottlenecks in Claude Flow swarms.

Official
Intermediate
QingHuan-6QingHuan-6

llm-torch-profiler-analysis

Triage torch-profiler traces into kernel, overlap, and fuse tables across LLM frameworks.

Community
Advanced
sqjiansqjian

sglang-torch-profiler-analysis

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

Community
Intermediate