Organize your machine learning experiments with ScalarStop

ScalarStop helps you train machine learning models by:

  • creating a system to uniquely name datasets, model architectures, trained models, and their hyperparameters.

  • saving and loading datasets and models to/from the filesystem in a consistent way.

  • recording dataset and model names, hyperparameters, and training metrics to a SQLite or PostgreSQL database.


ScalarStop is available on PyPI. You can install it from the command line using:

pip3 install scalarstop


The best way to learn ScalarStop is to follow the Official ScalarStop Tutorial.

Afterwards, you might want to dig deeper into the ScalarStop documentation. In general, a typical ScalarStop workflow involves four steps:

  1. Organize your datasets with scalarstop.datablob.

  2. Describe your machine learning model architectures using scalarstop.model_template.

  3. Load, train, and save machine learning models with scalarstop.model.

  4. Save hyperparameters and training metrics to a SQLite or PostgreSQL database using scalarstop.train_store.