FreeSkills
Commits, branches and sync with the but CLI
- 361361 downloads
- 198198 likes
- 103103 saves
Works with
- Claude Code
- Codex
- Cursor
- and 3 more
Benchmark inference and evaluate GPU kernels v1.0.0
Use this resource to measure LLM inference, analyze GPU kernels and compare their variants
New
- 312 downloads
- 216 likes
- 193 saves
Free
Install with the CLI
In a terminal, from the folder you want:
npx --yes palatekit@latest install llm-inference-benchmark-kernel-evalWith no option, the skill goes to ~/.agents/skills/; add --agent claude-code (or codex, cursor…) to target an agent.
First time: connect the CLI with
npx --yes palatekit@latest loginThe agent runs inference benchmarks (throughput, latency, TTFT, TPOT), captures PyTorch traces and analyzes them with TraceLens to find bottleneck kernels. It compares HIP, CUDA or Triton variants, then drives an optimization loop through end-to-end validation.
Mostly picked by
Updated November 2, 2026
Previews
Details
- Install guides
- Generates with
- Your agent's model
- Licence
- MIT
Similar resources
View all- Free
FreeSkills
Sentry CLI guide
- 525525 downloads
- 117117 likes
- 128128 saves
Works with- Claude Code
- Codex
- Cursor
- and 3 more
FreeSkills
Drizzle ORM best practices with PostgreSQL
- 407407 downloads
- 194194 likes
- 151151 saves
Works with- Claude Code
- Codex
- Cursor
- and 3 more
FreeSkills
SLO design wizard
- 537537 downloads
- 105105 likes
- 242242 saves
Works with- Claude Code
- Codex
- Cursor
- and 3 more
FreeSkills
User stories and sprint fit
- 451451 downloads
- 150150 likes
- 132132 saves
Works with- Claude Code
- Codex
- Cursor
- and 3 more
FreeSkills
Capacity-based sprint plan
- 423423 downloads
- 192192 likes
- 102102 saves
Works with- Claude Code
- Codex
- Cursor
- and 3 more
FreeSkills
Startup CTO advisor
- 580580 downloads
- 201201 likes
- 242242 saves
Works with- Claude Code
- Codex
- Cursor
- and 3 more
FreeSkills
Creating GitHub pull requests with quality checks
- 427427 downloads
- 230230 likes
- 196196 saves
Works with- Claude Code
- Codex
- Cursor
- and 3 more
What you get
- SKILL.md: choice of workflow (analyze, compare, benchmark) and preflight checks
- examples.md and reference.md: examples and command reference
Good fit for
- You want to measure TTFT and throughput of an inference engine
- You compare two implementations of the same kernel
Not for
Tags
- benchmark
- kernels
- inference
About
You say
“Compare these two vector_add kernel variants and rank them by performance.”
Skill published by amd under the MIT license, delivered in its original version.
Show moreShow less
How to use it
- Install the skill in your agent's skills folder
- Ask your agent to measure LLM inference, analyze GPU kernels and compare their variants
- Review and adjust the result
Content preview
Files
- .federated.json0.3 KB
- evals/evals.json5 KB
- evals/files/analyze-simple-hip/examples/simple_hip_test/analyze_default.yaml0.4 KB
- evals/files/analyze-simple-hip/examples/simple_hip_test/vector_add.hip1.8 KB
- evals/files/compare-hip-variants/baseline/vector_add.hip1.8 KB
- evals/files/compare-hip-variants/candidate/vector_add.hip1.9 KB
- examples.md5 KB
- LICENSE1.1 KB
- NOTICE-PALATEKIT.md0.5 KB
- reference.md10 KB
- skill-card.md0.5 KB
- SKILL.md9.3 KB
Origin and licence
Listed unmodified
- Author
- amd
- Source
- github.com/amd/skills/tree/0b64b9c4decd14af2e9789f1ef4c1b66909e68b1/skills/magpie-kernel-evaluator
- Licence
- MIT
Notice
« magpie-kernel-evaluator » (amd, MIT) livré dans sa version d'origine : fiche en français et en anglais par PalateKit.
User reviews
–
No ratings yet
Loading reviews…