Spines

Skeletons for parameterized models.

Important: This software is still in it’s early alpha phase and is
constantly in flux. It will likely change significantly.

Installation

To install spines use your package manager of choice, an example using
pipenv would be:

$ pipenv install spines

About

Spines is a library which provides a consistent (and hopefully familiar)
framework for building predictive models. It’s core Model class is
similar, in structure, to some of scikit-learn’s underlying Estimator
classes - but with a single set of unified functions for all models,
namely:

• Construct
• Fit
• Train
• Predict
• Error
• Score

The predict method is the only one that’s required to be
implemented, though the others are likely useful most of the time (and
often required to take advantage of some of the additional features
provided by spines).

Spines was built because the process of developing a model could be
significantly aided by an intelligent framework keeping tabs on changes,
storing results and helping you iterate. The purpose of spines was to
give a simple (and not too opinionated) interface/skeleton for models as
well as provide some helpful utilities for the model building process.
To accomplish this spines provides some useful key features:

• Standardized format for models of all types.
• Automatic version management.
• Storing intermediate/iterative results during the model development
  and training/fitting process.
• A unified storage format for models to facilitate collaboration,
  training and deployment.

Documentation

The latest documentation is hosted on
read the docs.

License

This project is licensed under the MIT License, for more information see
the LICENSE
file.