skills.rest
amd
amdOFFICIAL
Explore all skills by this author →
WebsiteGitHub

quark-onnx-autosearch-pro

Plans and orchestrates ONNX quantization auto-search runs with Optuna-driven presets and trial budgets.

154|27|Updated Jun 19, 2024
Search Tags:#calibration#quantization#model optimization#onnx#optuna#autosearch
One-click install
npx skills add https://github.com/amd/Quark --skill quark-onnx-autosearch-pro
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: quark-onnx-autosearch-pro
Source: https://github.com/amd/Quark/tree/main/.claude/skills-impl/l3-recipes/onnx/quark-onnx-autosearch-pro
Command: npx skills add https://github.com/amd/Quark --skill quark-onnx-autosearch-pro

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill removes the manual trial-and-error involved in finding a strong quantization setup for an ONNX model, especially when accuracy, latency, and deployment constraints all matter at once.

Core Features & Use Cases

  • End-to-end AutoSearchPro workflow: Orchestrates model intake, preset or custom search-space selection, calibration and evaluation reader setup, script generation, and confirmed execution.
  • Quantization tuning for ONNX models: Targets workflows such as choosing between ADVANCED_SEARCH, XINT8_SEARCH, A8W8_SEARCH, and A16W8_SEARCH, or validating a custom search space.
  • Safe search execution guardrails: Requires Optuna availability, checks device assumptions, bounds trial budgets, and avoids running the search directly before user approval.

Quick Start

Ask for a complete ONNX AutoSearchPro plan for my model, including the recommended preset, trial budget, generated script, and estimated cost before execution.

Frequently Asked Questions about quark-onnx-autosearch-pro

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

FAQPage Schema
How do I automate ONNX quantization tuning without manually setting search parameters?▼

Automating ONNX quantization uses an Optuna-driven auto-search to orchestrate model intake, calibration reader setup, and script generation. You select presets like ADVANCED_SEARCH or XINT8_SEARCH to bypass manual trial-and-error configuration.

What is the best way to choose between A8W8 and A16W8 quantization presets for an ONNX model?▼

Choosing between A8W8_SEARCH and A16W8_SEARCH presets involves validating your custom search space against accuracy and latency constraints. An auto-search workflow evaluates these presets using bounded trial budgets to find the optimal safe setup.

Can I use Optuna to auto-search ONNX quantization settings on specific hardware?▼

Yes, you can use Optuna to auto-search ONNX quantization settings, but the workflow checks device assumptions before execution. It requires device-aware configuration and explicit user approval to ensure the generated script runs safely on your target hardware.

How do I safely run a quantization search script without exceeding trial budgets?▼

To safely run a quantization search script, the auto-search workflow applies bounded trial budgets and resumable study artifacts. It generates a confirmed execution script only after explicit user approval, preventing unchecked resource consumption.

Why does my ONNX model quantization auto-search require explicit approval before execution?▼

ONNX model quantization auto-search requires explicit approval to enforce safe execution guardrails. This step ensures Optuna availability, validates device-aware configurations, and confirms bounded trial budgets before running the generated script.

Related Skills

Explore More Software Engineering Skills→
amdamd
154

quark-onnx-ptq

Execute end-to-end post-training quantization workflows for ONNX models with AMD Quark.

Official
Advanced
KameniAlexNeaKameniAlexNea

hpo

Run Bayesian hyperparameter optimization for LightGBM, XGBoost, and CatBoost models with Optuna.

Community
Intermediate
nahisahonahisaho
3

scientific-neural-architecture-search

Automate DARTS and Optuna neural-architecture search with Pareto optimization.

Community
Advanced
andikarachmanandikarachman
14

tuning-hyperparameters

Plan and execute hyperparameter tuning with Optuna and scikit-learn.

Community
Advanced