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.

About

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

Implementation of JAX JIT compilation strategies for hot path optimization
Vectorized operations using vmap to replace inefficient list comprehensions
Safe PRNG management through automated RNG key splitting patterns
JAX-compatible control flow using lax.scan and lax.cond for JIT compatibility
Scientific debugging patterns including NaN detection and immutable array handling
1 GitHub stars

Use Cases

Accelerating scientific simulations and numerical computations on GPU/TPU
Developing probabilistic models using Numpyro or Blackjax with idiomatic patterns
Refactoring standard NumPy code into high-performance JAX-native implementations

About

Key Features

Implementation of JAX JIT compilation strategies for hot path optimization
Vectorized operations using vmap to replace inefficient list comprehensions
Safe PRNG management through automated RNG key splitting patterns
JAX-compatible control flow using lax.scan and lax.cond for JIT compatibility
Scientific debugging patterns including NaN detection and immutable array handling
1 GitHub stars

Use Cases

Accelerating scientific simulations and numerical computations on GPU/TPU
Developing probabilistic models using Numpyro or Blackjax with idiomatic patterns
Refactoring standard NumPy code into high-performance JAX-native implementations