# patrick-kidger/jaxtyping

Type annotations and runtime checking for shape and dtype of JAX/NumPy/PyTorch/etc. arrays. https://docs.kidger.site/jaxtyping/

Repository: https://github.com/patrick-kidger/jaxtyping
Canonical: https://ross.abutalabs.com/products/jaxtyping
Language: Python
License: NOASSERTION
License Family: other
Topics: jax, python-typing, typing
Last push: 2026-08-18T21:40:02+00:00

## Health v2 (maintenance only)
Score: 95/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 98, release rhythm 88, longevity 100
- inputs: {"age_days": 1532, "days_push": 15, "days_rel": 81, "gap_med": 20, "n_releases_24m": 16}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1859, forks 94 (observed 2026-08-28T04:05:45.230632+00:00)

## What it is
A Python library providing type annotations and runtime type-checking for the shape and dtype of arrays and tensors in JAX, PyTorch, NumPy, MLX, and TensorFlow. It works alongside runtime checkers like typeguard or beartype.

## Use cases
- annotate tensor shapes and dtypes in function signatures
- runtime-check that arrays have expected shapes
- catch shape mismatch bugs in JAX or PyTorch code
- add shape annotations to NumPy array functions
- validate tensor dimensions in deep learning pipelines

## When to choose
- you write numerical or deep learning code in Python and want shape/dtype safety
- you use JAX, PyTorch, NumPy, MLX, or TensorFlow and want typed array APIs
- you want annotations compatible with typeguard or beartype

## When to avoid
- you need static type checking only without runtime verification
- your project does not involve array or tensor computations
- you need a framework-agnostic type system beyond array shapes

## Facets
- artifact type: library
- maturity: active
- function: type-system, testing, machine-learning
- domain: machine-learning, deep-learning, developer-tools, programming-languages
- platform: python, cross-platform
- tags: jax, pytorch, numpy, tensor-shapes, dtype-checking, runtime-type-checking, python-typing, annotations

## Member repositories
- patrick-kidger/jaxtyping (main) score 95

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:45.230632+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 2026-08-30T03:16:01.340753+00:00, confidence not recorded.
  - readme: https://github.com/patrick-kidger/jaxtyping (fetched 2026-08-28T04:05:45.230632+00:00, sha 3387a9f32625)
  - registry_pypi: https://pypi.org/pypi/jaxtyping/json (fetched 2026-08-29T10:55:15.296685+00:00, sha 924a7f8d68ef)
- Data as of 2026-08-30T08:39:29.467469+00:00.