Sacred

What it does

Command line

Its command-line interface can change parameters to run experiments with modified settings.checked 10 Oct 2026

Config access

Configuration parameters can be injected by name into functions.checked 4 Oct 2026

Configuration

Config scopes turn local variables in decorated functions into experiment configuration entries.checked 10 Oct 2026

Experiment parameters

Config scopes let local variables in a function define an experiment’s parameters.checked 4 Oct 2026

External dashboards

The README lists external frontends including Omniboard, Incense, and Sacredboard for viewing Sacred experiment data.checked 10 Oct 2026

Installation

The README says Sacred can be installed from PyPI using pip or installed from its Git repository.checked 10 Oct 2026

Integration

The documentation includes a dedicated integration section for TensorFlow.checked 10 Oct 2026

Integrations

The README describes a Neptune integration that logs Sacred experiments through an observer.checked 4 Oct 2026

License

The project is released under the MIT license.checked 10 Oct 2026

Limit

Sacred is designed for Python experiments; the README contrasts it with Sumatra for non-Python experiments.checked 4 Oct 2026

Optional dependencies

NumPy and PyMongo are optional dependencies that offer additional features.checked 10 Oct 2026

Parameter access

Captured functions can receive configuration entries through dependency injection.checked 10 Oct 2026

Purpose

Sacred helps configure, organize, log, and reproduce computational experiments.checked 10 Oct 2026

Reproducibility

Automatic seeding helps control randomness so experiment results remain reproducible.checked 10 Oct 2026

Run logging

Observers log experiment information and configuration and can save it to a database.checked 10 Oct 2026

Support

The README directs usage questions to Stack Overflow under the python-sacred tag and invites bug reports and feature requests through GitHub issues.checked 10 Oct 2026