JAX — autograd, JIT and XLA acceleration patterns
A scientific-agent reference for JAX — composable transformations (grad, jit, vmap, pmap), pure-function discipline, manual PRNG keys, critical rules, and CPU/GPU/TPU install patterns
- What
- A scientific-agent reference for JAX — composable transformations (grad, jit, vmap, pmap), pure-function discipline, manual PRNG keys, critical rules, and CPU/GPU/TPU install patterns
- Cost
- Free
- Needs
- Python 3.9+; for GPU, a CUDA-compatible card and the matching jaxlib CUDA build (check official docs); for CPU, plain pip works
- Install
- Copy the installer prompt below into your Muse — your agent does the rest.
Curated by Skill Harbor — @tondevrel's jax skill, listed here with credit to its creator: a scientific-agent reference for JAX — Google's framework combining a NumPy-like API with composable function transformations (grad for differentiation, jit for XLA compilation, vmap for vectorization, pmap for parallelization) — covering when to use it (high-performance scientific simulations, higher-order derivatives, physics-informed ML, differentiable simulations, CPU/GPU/TPU portability), the core principles (pure functions and immutability, manual PRNG key management, XLA compilation), a quick reference (install patterns for CPU and CUDA, standard imports, the differentiate-and-JIT basic pattern), and critical do/don't rules for writing correct JAX code. Honest caveats: it's a reference playbook for the upstream google/jax project — install the actual jax/jaxlib from the official releases (match your CUDA version for GPU) and read the official docs at jax.readthedocs.io; JAX's functional discipline (pure functions, explicit random keys) is a real mental-model shift from NumPy/PyTorch — expect a learning curve; GPU/TPU acceleration is where the value is — on plain CPU, PyTorch or NumPy may be simpler. MIT licensed. Skill Harbor never reviews the code, review it yourself before use. Discovered via skills.sh.
Version:
Install
Prerequisites: Python 3.9+; for GPU, a CUDA-compatible card and the matching jaxlib CUDA build (check official docs); for CPU, plain pip works Install "JAX — autograd, JIT and XLA acceleration patterns" for me. It gives my agent @tondevrel's JAX reference: composable transformations (grad, jit, vmap, pmap), pure-function discipline, manual PRNG keys, the differentiate-and-JIT pattern, install recipes for CPU/GPU/TPU, and critical rules for correct JAX code. MIT licensed. Repository: https://github.com/tondevrel/scientific-agent-skills/blob/main/skills/jax/SKILL.md 1. Fetch the SKILL.md file (and any helper files) from the repository path into a temporary folder and summarize what it does in one or two sentences. 2. Safety check: review the SKILL.md and scripts for anything suspicious (unexpected network calls, shell commands, credential harvesting). Install commands should point only at official jax releases — any third-party wheel index is a red flag. This repo should contain zero secrets in code. Verify that holds here; STOP on any red flag and tell me. 3. Install it as a skill: copy SKILL.md and its helper files into the agent's skills directory, in a folder named "jax". 4. Verify with no network calls: frontmatter valid, files in place. 5. Report what was installed, where, and what I still need to do myself (e.g. install jax/jaxlib myself from the official releases matching my CUDA version, read the official docs at jax.readthedocs.io, and budget learning time for the pure-function discipline). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Never require it unless it's in the prerequisites above. Rules: don't touch anything outside the temp folder and the install target. If anything looks off, stop and ask me.
Questions
How do I install a build?
Every product page includes a copy-paste install prompt. Paste it into your Muse and it sets the build up for you — no manual configuration.
Where does my money go?
Straight to the seller. Skill Harbor never processes payments: checkout happens on the seller’s own page, usually Stripe.
What does the ✓ next to a creator’s name mean?
It means we confirmed the identity of the person behind the listing. It says nothing about the code itself — always check a build before installing it.