jax

JAX high-performance numerical computing. Use for ML research.

7 stars

byG1Joshi

View on GitHub Installation ↓

Best use case

jax is best used when you need a repeatable AI agent workflow instead of a one-off prompt.

JAX high-performance numerical computing. Use for ML research.

Teams using jax should expect a more consistent output, faster repeated execution, less prompt rewriting.

When to use this skill

You want a reusable workflow that can be run more than once with consistent structure.

When not to use this skill

You only need a quick one-off answer and do not need a reusable workflow.
You cannot install or maintain the underlying files, dependencies, or repository context.

Installation

Claude Code / Cursor / Codex

$curl -o ~/.claude/skills/jax/SKILL.md --create-dirs "https://raw.githubusercontent.com/G1Joshi/Agent-Skills/main/skills/ai-ml/jax/SKILL.md"

Manual Installation

Download SKILL.md from GitHub
Place it in .claude/skills/jax/SKILL.md inside your project
Restart your AI agent — it will auto-discover the skill

How jax Compares

Feature / Agent	jax	Standard Approach
Platform Support	Not specified	Limited / Varies
Context Awareness	High	Baseline
Installation Complexity	Unknown	N/A

Frequently Asked Questions

What does this skill do?

JAX high-performance numerical computing. Use for ML research.

Where can I find the source code?

You can find the source code on GitHub using the link provided at the top of the page.

SKILL.md Source

# JAX

JAX is "NumPy on steroids". It combines Autograd (automatic differentiation) with XLA (compilation). 2025 sees **Flax NNX** (PyTorch-style OOP) becoming standard.

## When to Use

- **TPU Training**: JAX runs natively on Google TPUs.
- **Research**: If you need to compute 10th order derivatives or strange math.
- **Massive Scale**: DeepMind and OpenAI use JAX for training frontier models.

## Core Concepts

### Functional Transformations

`grad()`, `jit()`, `vmap()`, `pmap()`.

### Flax (NNX)

Neural network library. NNX introduces mutable state (OOP) to make JAX feel like PyTorch.

### Statelessness

(Legacy Flax) parameters are stored separately from the model.

## Best Practices (2025)

**Do**:

- **Use `jit`**: Always compile your functions.
- **Use Flax NNX**: Avoid the complexity of legacy immutable Flax/Haiku.
- **Use `shard_map`**: For distributed training across devices.

**Don't**:

- **Don't use side effects**: `print()` inside a `jit` function only runs once (during tracing).

## References

- [JAX Documentation](https://jax.readthedocs.io/)