ptq-workflow-integration

Integrate a new PTQ workflow into cache-dit with a public API and validation.

1.2k|77|Updated Jun 12, 2025
One-click install
npx skills add https://github.com/vipshop/cache-dit --skill ptq-workflow-integration
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: ptq-workflow-integration
Source: https://github.com/vipshop/cache-dit/tree/main/.copilot/skills/ptq-workflow-integration
Command: npx skills add https://github.com/vipshop/cache-dit --skill ptq-workflow-integration

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill provides a structured approach to integrating a new PTQ workflow into cache-dit, ensuring that the public API, validation logic, and save/load semantics are thoughtfully designed and maintainable.

Core Features & Use Cases

  • Establishes a centralized design pattern for PTQ integrations, including config schema placement, backend orchestration, and deterministic serialization.
  • Guides the team through planning, implementing, testing, and documenting a PTQ backend so that new algorithms can be added with predictable UX and minimal boilerplate.
  • Aligns with repo style guidelines, ensuring public APIs remain stable and future PTQ work stays within cache-dit's architectural boundaries.

Quick Start

Define your PTQ backend, sketch the public API surface, and begin a small-scale integration plan to validate save/load paths and test coverage.

Frequently Asked Questions about ptq-workflow-integration

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

FAQPage Schema
How do I integrate a new PTQ workflow into cache-dit?

Integrate a PTQ workflow by defining the backend, sketching the public API surface, and validating save/load paths. This approach enforces centralized validation, modular design, and repo-relative references to ensure consistent UX and test coverage.

What is the best way to design a public API for post-training quantization backends?

The best way to design a public API for post-training quantization backends is through centralized validation and deterministic serialization. Aligning with repo style guidelines keeps APIs stable and within architectural boundaries for scalable implementations.

How does cache-dit handle save and load operations for quantization artifacts?

Cache-dit handles save and load operations for quantization artifacts using deterministic serialization paths. This ensures reliable artifact storage and retrieval across backend integrations and calibration flows with robust error handling.

Does this PTQ integration approach support scalable implementations across different backends?

Yes, this PTQ integration approach supports scalable implementations across different backends. Enforcing modular design and centralized validation allows new algorithms to be added with predictable UX and minimal boilerplate.

Why do I need centralized validation for PTQ calibration flows?

Centralized validation is needed for PTQ calibration flows to ensure consistent UX and robust error handling. It maintains architectural boundaries and supports reliable serialization paths across backend integrations.