rail.core.model module

Wrapper class for ML Models

class rail.core.model.Model

Bases: object

Class to act as wrapper for ML models

This will attach metadata to the ML model, and provide tools for validation, versioning and tracking provenance.

__init__(data, creation_class_name, version=0, catalog_tag=None, provenance=None)

Constructor

Parameters:

  • data (Any) – Model data

  • creation_class_name (str) – Name of class that created this model

  • version (int) – Version of the model

  • catalog_tag (str | None) – Associated CatalogConfigBase that defined training dataset

  • provenance (dict | None) – Provenance infomration

Return type:

None

classmethod dump(obj, path, creation_class_name='dummy', version=0, catalog_tag=None, provenance=None)

Write an object to a model file

This will promote it to a Model if it isn’t already

Parameters:

  • obj (Any) – Object to dump

  • path (str) – File to write

  • creation_class_name (str) – Name of class that created this model

  • version (int) – Version of the model

  • catalog_tag (str | None) – Associated CatalogConfigBase that defined training dataset

  • provenance (dict | None) – Provenance information

Returns:

Newly converted Model

Return type:

Model

classmethod read(path, creation_class_name='dummy', version=0, catalog_tag=None, provenance=None)

Read a model from a file.

Note that this will promote the data to a Model if it is not already

Parameters:

  • path (str) – File to read

  • creation_class_name (str) – Name of class that created this model

  • version (int) – Version of the model

  • catalog_tag (str | None) – Associated CatalogConfigBase that defined training dataset

  • provenance (dict | None) – Provenance infomration

Returns:

Newly read Model

Return type:

Model

validate(creation_class_name, version)
Parameters:

  • creation_class_name (str | None) – Name of class that created this model

  • version (int | None) – Version of the model

Return type:

None

:raises TypeError : Either creation_class_name or version does not match:

classmethod wrap(inpath, outpath, creation_class_name='dummy', version=0, catalog_tag=None, provenance=None)

Read a model from a file and write it as a Model if it is not already

Parameters:

  • inpath (str) – File to read

  • outpath (str) – File to write

  • creation_class_name (str) – Name of class that created this model

  • version (int) – Version of the model

  • catalog_tag (str | None) – Associated CatalogConfigBase that defined training dataset

  • provenance (dict | None) – Provenance information

Returns:

Newly read & converted Model

Return type:

Model

write(path)

Write a model to a file

Parameters:

path (str) – File to write

Return type:

None