All Tools
S
MonitoringFreeOpen Source
SACRED
Experiment tracking and configuration for ML
MIT
ABOUT
ML experiments quickly become chaotic without systematic tracking of hyperparameters, code versions, outputs, and hardware configurations. Sacred provides a lightweight, decorator-based framework that automatically captures experiment configurations, source code snapshots, dependencies, metrics, and outputs — with built-in support for MongoDB and SQLite backends — making every experiment reproducible with a single command, even across different machines and team members.
INSTALL
pip install sacredINTEGRATION GUIDE
1. Track hyperparameter configurations and training metrics across hundreds of model experiments with automatic logging
2. Reproduce any past experiment exactly by replaying its captured configuration, source code, and dependency snapshot
3. Organize team experiments in a shared MongoDB database with searchable experiment records and run comparisons
4. Integrate experiment logging into existing training scripts with minimal boilerplate using Python decorators
TAGS
experiment-trackingreproducibilitypythonconfigurationloggingml-research