---
id: sweeps
version: "0.2.0"
license: MIT license
license_treatment: permissive
maintenance: active
---

# sweeps — Weights and Biases Hyperparameter Sweeps Engine.

License: permissive · Maintenance: active · Downloads: 103.1K/mo

## What it is and what it does

Sweeps is a hyperparameter optimization engine that powers Weights & Biases' sweep functionality. It takes a sweep configuration (defining the optimization method, parameter ranges, and metric to optimize) and a list of completed runs, then suggests the next hyperparameter values to try. It supports multiple search strategies including Bayesian optimization and grid search, and can identify runs to stop early using methods like Hyperband to save compute resources.

The package validates configurations against a JSON schema, routes to the appropriate search algorithm, and returns structured suggestions. It's designed as a backend component but is also available as a standalone library for local hyperparameter tuning workflows. Dependencies include numpy, scipy, scikit-learn for numerical computation, and pydantic for configuration validation.

Use it for:

- Suggest next hyperparameters in a machine learning training loop using Bayesian optimization to balance exploration and exploitation.
- Implement early stopping in parallel hyperparameter searches by identifying underperforming runs via Hyperband.
- Validate and parse sweep configurations to ensure they conform to the expected schema before running experiments.
- Integrate hyperparameter optimization into a local training pipeline without relying on external services.
- Grid search over discrete parameter combinations with deterministic parameter assignment.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Generates hyperparameter sweep suggestions using Bayesian optimization, grid search, and other methods, and identifies runs to early-terminate based on performance.

Yes. The package is actively maintained, has no known vulnerabilities, carries a permissive MIT license, and offers low-friction installation. It solves a concrete problem—suggesting hyperparameters and early-stopping decisions—with established algorithms. Install it if you need programmatic hyperparameter optimization or are integrating with Weights & Biases sweeps locally.

## Install

pip install sweeps
uv add sweeps
poetry add sweeps

## Installing sweeps

Before you install: Low friction install with a pure-Python wheel. Actively maintained as of August 2026 with no known vulnerabilities. Depends on established scientific libraries (numpy, scipy, scikit-learn, pydantic).

License in practice: MIT license permits commercial and private use with minimal restrictions; you may use, modify, and distribute freely provided you include the license notice.

Quickstart:

pip install sweeps

from sweeps import next_run, SweepRun, RunState

config = {
    "metric": {"name": "loss", "goal": "minimize"},
    "method": "bayes",
    "parameters": {"v1": {"min": 1, "max": 10}},
}
runs = []
suggestion = next_run(config, runs)

Verify before relying:

- Whether the package is actively developed beyond August 2026 or if maintenance has stalled since the last release in August 2022.
- Performance characteristics and scalability limits for large numbers of runs or high-dimensional parameter spaces.
- Availability of documentation beyond the README examples provided.

## Package facts

- License: MIT license (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 103.1K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

hyperparameter optimization, bayesian hyperparameter search, hyperparameter sweep suggestions, early stopping hyperband, grid search hyperparameters, automated parameter tuning, hyperparameter tuning engine, hyperparameter-optimization, machine-learning, bayesian-search

[View on SkillFed](https://skillfed.io/packages/sweeps) · [View on PyPI](https://pypi.org/project/sweeps/)