add-pallas-kernel

Official

Optimize TPU/GPU kernels with Pallas.

Authormarin-community
Version1.0.0
Installs0

System Documentation

What problem does it solve?

This Skill addresses the need to implement, modify, benchmark, or autotune high-performance kernels on TPUs and GPUs using JAX's Pallas experimental features, ensuring numerical safety and performance.

Core Features & Use Cases

  • Kernel Implementation: Develop both vanilla JAX and Pallas kernel implementations with consistent APIs.
  • Correctness & Performance Validation: Includes harnesses for value/gradient parity, numerics, and steady-state timing.
  • Autotuning: Automates the search for optimal block/tile sizes for specific hardware and shape regimes.
  • Use Case: When developing a new neural network layer that requires custom, high-performance computation on accelerators, this Skill provides the framework to build, test, and optimize the underlying kernel.

Quick Start

Implement a Pallas kernel for matrix multiplication with autotuned block sizes for TPU v5e.

Dependency Matrix

Required Modules

None required

Components

references

💻 Claude Code Installation

Recommended: Let Claude install automatically. Simply copy and paste the text below to Claude Code.

Please help me install this Skill:
Name: add-pallas-kernel
Download link: https://github.com/marin-community/marin/archive/main.zip#add-pallas-kernel

Please download this .zip file, extract it, and install it in the .claude/skills/ directory.
View Source Repository

Agent Skills Search Helper

Install a tiny helper to your Agent, search and equip skill from 223,000+ vetted skills library on demand.