# HIPS/autograd

Efficiently computes derivatives of NumPy code.

Repository: https://github.com/HIPS/autograd
Canonical: https://ross.abutalabs.com/products/autograd
Language: Python
License: MIT
License Family: permissive
Topics: autograd, automatic-differentiation, numpy, python, deep-learning, derivative, jax, machine-learning, neural-network, numpy-arrays, scipy, ufunc, backpropagation, differentiation
Last push: 2026-08-24T18:24:36+00:00

## Health v2 (maintenance only)
Score: 85/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 58, longevity 100
- inputs: {"age_days": 4300, "days_push": 9, "days_rel": 65, "gap_med": 210.5, "n_releases_24m": 3}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 7523, forks 940 (observed 2026-08-28T04:10:00.483380+00:00)

## What it is
Autograd is a Python library that efficiently computes derivatives of native Python and NumPy code, supporting reverse-mode (backpropagation) and forward-mode automatic differentiation. It handles a large subset of Python features including loops, ifs, recursion, and closures, and supports higher-order derivatives.

## Use cases
- compute gradients of numpy functions for gradient-based optimization
- take derivatives of derivatives of a function
- backpropagate through a neural network implemented in numpy
- differentiate scipy functions
- compute elementwise derivatives of vectorized functions
- verify gradients against finite differences

## When to choose
- you need gradients of plain Python/NumPy code without rewriting it in a deep learning framework
- you want a lightweight, minimal autodiff library for research or teaching
- you need higher-order derivatives or mixed forward/reverse-mode differentiation

## When to avoid
- you need GPU acceleration or large-scale deep learning training - use JAX or PyTorch instead
- you need a full deep learning framework with optimized kernels and distributed training
- your code relies on operations outside the supported NumPy/SciPy subset

## Facets
- artifact type: library
- maturity: stable
- function: machine-learning, math
- domain: machine-learning, data-science
- platform: python
- tags: automatic-differentiation, numpy, backpropagation, gradients, optimization, scipy, algorithms

## Member repositories
- HIPS/autograd (main) score 85

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:00.483380+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 2026-08-29T17:37:28.425270+00:00, confidence not recorded.
  - readme: https://github.com/HIPS/autograd (fetched 2026-08-28T04:10:00.483380+00:00, sha 0d5c88adb29d)
  - registry_pypi: https://pypi.org/pypi/autograd/json (fetched 2026-08-29T08:32:33.153934+00:00, sha cb1e7ceb6c91)
- Data as of 2026-08-30T08:39:29.467469+00:00.