Skip to content
Back to skills

Validation

ASecurity

"Evaluates exported MMDeploy models, profiles latency, runs

  • 247 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added September 8, 2026
devopsrailsbackend

Security analysis

A100/100

Pro scans all 10 files and shows the line behind each finding

Scanned September 8, 2026

npx -y skills add VectorSpaceLab/AREX-Skill --skill validation --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Validation?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Validation
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vectorspacelab-validation/badge)](https://www.skillsdirectory.com/skills/vectorspacelab-validation)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: validation
description: "Evaluates exported MMDeploy models, profiles latency, runs
  regression matrices, and generates backend coverage tables."
disable-model-invocation: true
metadata:
  disco-role: operating
license: Apache 2.0
---

# Validation

Use this sub-skill when a task asks to evaluate an exported MMDeploy backend model, measure inference speed, run a regression matrix, build a supported-backend table, or interpret validation/profiling reports.

## Route by task

- **Metric-aware backend evaluation:** read [evaluation](references/evaluation.md), then use [scripts/test.py](scripts/test.py) when the user needs dataset metrics, visualized outputs, speed-test logging during evaluation, backend model-file selection, or `--cfg-options` overrides.
- **Latency-only profiling:** read [profiling](references/profiling.md), then use [scripts/profiler.py](scripts/profiler.py) when the user asks for warmup/iteration-controlled latency or custom batch-size/shape timing without evaluator metrics.
- **Maintainer regression matrix:** read [regression](references/regression.md), then use [scripts/regression_test.py](scripts/regression_test.py) for codebase/backend/model filters, conversion-only versus precision modes, checkpoint caching, and merged report workbooks.
- **Supported model/backend table:** read [supported model tables](references/supported-model-tables.md), then use [scripts/generate_md_table.py](scripts/generate_md_table.py) to turn regression matrix YAML into Markdown support tables.
- **Unexpected failure, skip, or slow run:** read [troubleshooting](references/troubleshooting.md) before classifying a problem as a model defect.

## Guardrails

- Use `test.py` for validation/evaluation and optional speed logs; use `profiler.py` for latency-only measurements. Do not treat profiler output as metric evidence.
- Regression runs can download checkpoints, call conversion workflows, and exercise heavyweight backends. Always narrow `--codebase`, `--backends`, and `--models` unless the user explicitly requests a full matrix.
- Backend installation/build problems belong to the backend workflow. In this validation sub-skill, record whether a backend is unavailable, skipped, failed, or blocked, and route installation remediation elsewhere.
- Graph rewrite or optimizer implementation changes belong to the extensibility workflow. Use validation here only to observe downstream metric/latency/report impact.

Files in this skill

  • SKILL.md2.4 KB
  • references/evaluation.md5.8 KB
  • references/profiling.md3.9 KB
  • references/regression.md7.3 KB
  • references/supported-model-tables.md3.3 KB
  • references/troubleshooting.md8.4 KB
  • scripts/generate_md_table.py3.2 KB
  • scripts/profiler.py6.2 KB
  • scripts/regression_test.py40.6 KB
  • scripts/test.py5.8 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.