Does this skill handle combined type groups?

Yes, it correctly maps combined patterns like AT_DISPATCH_ALL_TYPES_AND_COMPLEX by using multiple AT_EXPAND calls within the V2 macro.

What is AT_DISPATCH_V2?

AT_DISPATCH_V2 is the modernized dispatch API in PyTorch that replaces fixed-arity macros with a more composable and extensible format, allowing for easier type management.

Will this skill modify my existing header includes?

The skill adds the required #include header while preserving the original Dispatch.h to ensure other parts of the file remain functional.

Can I use this for CUDA kernels?

Absolutely. This skill is specifically designed for files in the ATen native directory, including CUDA kernels and native C++ operator implementations.

Why is AT_WRAP required in the conversion?

AT_WRAP is mandatory in the V2 API to safely wrap lambda expressions that contain internal commas, preventing C++ preprocessor parsing errors.

PyTorch Dispatch V2 Migrator

Name: PyTorch Dispatch V2 Migrator
Author: aiskillstore

byaiskillstore

•

179

•

Data Science & ML

Automates the migration of legacy PyTorch AT_DISPATCH macros to the modernized AT_DISPATCH_V2 API in C++ kernel implementations.

This skill streamlines the refactoring of PyTorch ATen C++ code by converting legacy dispatch macros to the V2 format defined in Dispatch_v2.h. It handles complex argument reordering, applies mandatory AT_WRAP and AT_EXPAND patterns, and ensures accurate type group mapping for macros like AT_ALL_TYPES or AT_FLOATING_TYPES. It is an essential tool for framework developers and contributors maintaining or porting PyTorch kernels, CUDA implementations, and native operators to ensure compatibility with the latest extensible dispatch standards.

Key Features

01Automatic header management for Dispatch_v2.h

02Automated argument reordering for V2 API compliance

03Mandatory AT_WRAP integration for complex lambda functions

04179 GitHub stars

05Intelligent mapping of legacy type macros to V2 type groups

06Support for multi-line lambdas and high-arity patterns (AND2, AND3, etc.)

Use Cases

01Refactoring CUDA kernel implementations for improved type extensibility

02Standardizing C++ native operator code by removing arity-specific macro names

03Porting legacy ATen kernels to the modernized PyTorch dispatch system

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add aiskillstore/marketplace pytorch

For use in Claude.ai and ChatGPT

Download Skill