# patrick-kidger/equinox

Elegant easy-to-use neural networks + scientific computing in JAX. https://docs.kidger.site/equinox/

Repository: https://github.com/patrick-kidger/equinox
Canonical: https://ross.abutalabs.com/products/equinox
Language: Python
License: Apache-2.0
License Family: permissive
Topics: deep-learning, jax, neural-networks, equinox
Last push: 2026-08-10T10:52:15+00:00

## Health v2 (maintenance only)
Score: 92/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 97, release rhythm 82, longevity 100
- inputs: {"age_days": 1862, "days_push": 23, "days_rel": 120, "gap_med": 29.5, "n_releases_24m": 19}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2957, forks 211 (observed 2026-08-28T04:07:32.237640+00:00)

## What it is
Equinox is a Python library providing neural networks and scientific computing utilities for JAX, using PyTorch-like class-based syntax where models are registered as PyTrees. It includes filtered API transformations, PyTree manipulation routines, and advanced features like runtime errors, all fully interoperable with the rest of the JAX ecosystem.

## Use cases
- build neural networks in JAX with PyTorch-like syntax
- define models that pass smoothly across jit/grad/vmap boundaries
- manipulate PyTrees and filter parameters in JAX transformations
- train CNNs or other models on datasets like MNIST
- replace Flax or Haiku with a simpler model-building approach
- do scientific computing and differentiable programming in JAX

## When to choose
- you use JAX and want an elegant, PyTorch-like way to define models
- you need models that are plain PyTrees compatible with all JAX transformations
- you want filtered transformations and advanced features like runtime errors
- you find Flax or Haiku too complex for your needs

## When to avoid
- you are not using JAX (e.g. you want PyTorch or TensorFlow)
- you need a full training framework with built-in trainers and data loaders
- you rely on ecosystem tooling specific to Flax or Haiku

## Facets
- artifact type: library
- maturity: stable
- function: machine-learning, deep-learning, llm-training
- domain: deep-learning, machine-learning, artificial-intelligence, data-science
- platform: python, cross-platform
- tags: jax, neural-networks, pytrees, pytorch-like-api, scientific-computing, filtered-transformations, gpu

## Member repositories
- patrick-kidger/equinox (main) score 92

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:32.237640+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:47:31.968027+00:00, confidence not recorded.
  - readme: https://github.com/patrick-kidger/equinox (fetched 2026-08-28T04:07:32.237640+00:00, sha 668134eae972)
  - registry_pypi: https://pypi.org/pypi/equinox/json (fetched 2026-08-29T09:47:32.730471+00:00, sha 79fd50b6fce1)
- Data as of 2026-08-30T08:39:29.467469+00:00.