# facebookresearch/nevergrad

A Python toolbox for performing gradient-free optimization

Repository: https://github.com/facebookresearch/nevergrad
Canonical: https://ross.abutalabs.com/products/nevergrad
Homepage: https://facebookresearch.github.io/nevergrad/
Language: Python
License: MIT
License Family: permissive
Last push: 2026-07-24T21:41:13+00:00

## Health v2 (maintenance only)
Score: 76/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 94, release rhythm 40, longevity 100
- inputs: {"age_days": 2843, "days_push": 40, "days_rel": 497, "gap_med": 14.5, "n_releases_24m": 7}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 4204, forks 371 (observed 2026-08-28T04:08:38.993976+00:00)

## What it is
Nevergrad is a Python toolbox for gradient-free (derivative-free) optimization, providing many optimization algorithms such as NGOpt and differential evolution. It supports optimizing continuous, discrete, and mixed parameter spaces, making it well suited for hyperparameter tuning and black-box function minimization.

## Use cases
- tune machine learning hyperparameters without gradients
- minimize a black-box function with no derivative access
- optimize mixed continuous and discrete parameters
- run evolutionary algorithms like differential evolution in Python
- benchmark optimization algorithms
- optimize simulation or experiment parameters

## When to choose
- your objective function is non-differentiable, noisy, or expensive to differentiate
- you need to tune hyperparameters of ML models or pipelines
- you need a variety of derivative-free optimizers behind one API
- your search space mixes continuous, integer, and categorical variables

## When to avoid
- your problem is smooth and gradient-based methods (e.g., Adam, L-BFGS) work well
- you need large-scale deep learning training with backpropagation
- you need a general-purpose constraint solver rather than black-box optimization

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, benchmarking, data-science
- domain: machine-learning, artificial-intelligence
- platform: python, cross-platform
- tags: gradient-free-optimization, derivative-free-optimization, hyperparameter-tuning, evolutionary-algorithms, metaheuristic, algorithms

## Member repositories
- facebookresearch/nevergrad (main) score 76

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:38.993976+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 2026-08-29T18:22:31.832267+00:00, confidence not recorded.
  - readme: https://github.com/facebookresearch/nevergrad (fetched 2026-08-28T04:08:38.993976+00:00, sha 2fd7676d6a2c)
  - homepage: https://facebookresearch.github.io/nevergrad/ (fetched 2026-08-29T09:12:46.735922+00:00, sha fdb12ce39e7b)
  - registry_pypi: https://pypi.org/pypi/nevergrad/json (fetched 2026-08-29T09:12:46.744961+00:00, sha 507b1d118246)
- Data as of 2026-08-30T08:39:29.467469+00:00.