How does it handle JAX's immutable arrays?

The skill provides patterns for functional updates and checks for common errors related to JAX array immutability and side effects.

How does this skill improve JAX code performance?

It identifies hot paths for @jax.jit application and replaces standard Python loops with jax.vmap or jax.lax.scan to leverage hardware acceleration.

Does it help with JAX random number generation?

Yes, it enforces the 'always split' rule for RNG keys to ensure that random states are handled correctly and predictably across functions.

Can this skill help with libraries like Numpyro?

Absolutely. It is designed to support the broader JAX ecosystem, including Numpyro, Distrax, and Blackjax for probabilistic programming.

Why use lax.scan instead of standard Python loops?

Standard Python loops are unrolled during JIT compilation, which can lead to long compile times. lax.scan keeps the loop structure for more efficient compilation.

JAX Scientific Computing

Name: JAX Scientific Computing
Author: yallup

byyallup

•

Data Science & ML

Applies idiomatic JAX patterns and best practices to optimize scientific computing and machine learning workflows in Python.

This Claude Code Skill provides specialized guidance for developing high-performance scientific applications using JAX and its ecosystem, including Distrax, Numpyro, and Blackjax. It enforces essential JAX principles such as functional purity, efficient JIT compilation, and vectorization via vmap to replace standard Python loops. By providing patterns for proper PRNG key management and JAX-native control flow (lax.scan, lax.cond), this skill helps developers avoid common pitfalls like side effects and performance bottlenecks in accelerated computing environments.

Key Features

01Implementation of JAX JIT compilation strategies for hot path optimization

02Vectorized operations using vmap to replace inefficient list comprehensions

03Safe PRNG management through automated RNG key splitting patterns

04JAX-compatible control flow using lax.scan and lax.cond for JIT compatibility

05Scientific debugging patterns including NaN detection and immutable array handling

061 GitHub stars

Use Cases

01Accelerating scientific simulations and numerical computations on GPU/TPU

02Developing probabilistic models using Numpyro or Blackjax with idiomatic patterns

03Refactoring standard NumPy code into high-performance JAX-native implementations

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yallup/claude-skills using-jax

For use in Claude.ai and ChatGPT

Download Skill