Export and import generators

MOSTLY AI allows you to export already trained generators from one environment and import them into another.

The export and import of generators is useful when you need to have a separation of concerns. For example, you can train a generator with the Python SDK in an environment where you have direct and secure access to the original data, and where it is safe to train generators on the original data. In this scenario, the data remains private as the generator learns only the statistical properties and never memorizes specific data points.

You can then export the trained generator from the secure environment and import it into a running instance of the MOSTLY AI Platform. That instance can be made available to everyone in your organization. People can then use the Platform to generate privacy-safe synthetic data and extract data insights with the Assistant without access to the original data.

Export a generator

If you use the UI of the MOSTLY AI Platform, you can export a generator in one of two ways.

Prerequisites

To export a generator, you must be its Owner or have the Editor role if the generator was shared with you.

Steps

Export a generator directly from the Generators page.
1. From the Generators page, click the kebab menu of a generator, and select Export to file.
Export a generator after you open it.
1. From the Generators page, click a generator to open it.
2. Click the kebab menu in the upper right, and select Export to file.

If you trained a generator in a local environment with the Python SDK, you can export it to a ZIP archive with the export_to_file() method.

Optionally, you can specify a filename for the exported generator.

python

g = mostly.generators.get('INSERT_GENERATOR_ID')
 
# Exports with a default filename 'generator-UUID-time-high-octet.zip' 
g.export_to_file() 
 
# Export with a specified filename
g.export_to_file('INSERT_GENERATOR_FILENAME.zip')

If you use the Python SDK with a remote Platform instance, you can export a generator from the Platform. Use the same export_to_file() method as you would in a local environment.

python

g = mostly_remote.generators.get('INSERT_GENERATOR_ID')
 
# Exports with a default filename 'generator-UUID-time-high-octet.zip' 
g.export_to_file() 
 
# Export with a specified filename
g.export_to_file('INSERT_GENERATOR_FILENAME.zip')

Result

The generator is exported as a ZIP archive. You can now import it into another instance of the MOSTLY AI Platform (local or remote).

Import a generator

Learn how to import generators into another instance of the MOSTLY AI Platform with the UI and the Python SDK. You can do so in a local environment as well as in a remotely running MOSTLY AI Platform.

You can use the UI of another instance of the MOSTLY AI Platform to import exported generators.

Steps

On the Generators page, click Import from file.
On the Import Generator prompt, click in the upload area to browse or drag-and-drop the exported generator ZIP file.
Click Proceed.

Result

The imported generator opens immediately. You can now use it to generate synthetic data in the new environment.

If you use the Python SDK in a local environment, you can import a generator with the mostly.generators.import_from_file method.

python

g = mostly.generators.import_from_file('path/to/generator.zip')

If you use the Python SDK with a remote Platform instance, you can again use the mostly_remote.generators.import_from_file method to import a generator.

python

g = mostly_remote.generators.import_from_file('path/to/generator.zip')

Generator folder structure

You can explore the interactive folder and file structure of exported generators below.

📑

The structure below is an example containing template filenames.

For example, <table_name> is replaced by the actual table name in the generator.

In an actual exported generator:

the template filenames depend on model types and table names
the number of model folders and QA reports depend on the number of models and the number of tables in the generator.

generator.json
job_progress.json