Altair SmartWorks Analytics

 

Model Registry

The Model Registry tab is one of the two tabs that displays when the MLOps node is opended. It displays a database of models that have been registered with the application.

 

The tab displays several columns

  • Model Name

  • Model Type – “SKLEARN“ for scikit-learn models, “KSTUDIO“ for Knowledge Studio models, or “PYTHON“ for Python-code models. The displayed type is that of the latest version of the model

  • Latest Version – the latest registered version for that particular model

  • Created by – the user who created that model

  • Created Time – when this model was first registered with version 1

  • Last Modified Time – the last time a the model was updated (with a new version)

You can sort any of these columns by clicking on the sorting buttons located to the right of each column header. Typing the first few characters of a model name into the Search box located at the upper right-hand corner of the page displays all models corresponding to this string. Models in different MLFlow repositories can be viewed by clicking and selecting the appropriate repository from the drop-down list located beside the Search bar.

You can also register a new model version by clicking on the Register New Model button.

Model Viewer Page

The Model Viewer page comes up when a user clicks into a particular model.

 

This page shows the following columns

  • Created by – the user who created that model

  • Registered at – the time at which this model version was registered

  • Model Version – the version for a particular model

  • Model Type – “SKLEARN“ for scikit-learn models, “KSTUDIO“ for Knowledge Studio models, or “PYTHON“ for Python-code models. The displayed type is that of the latest version of the model

  • Package Version – the version of “Sklearn“ or “Knowledge Studio“ or “Python“ this model version has been registered under

You can sort any of these columns by clicking on the sorting buttons located to the right of each column header.

You can also register a new model version by clicking on the Add new model version button.

Registering New Models

You can register new models via the Model Registry page.

Steps:

  1. Click Register New Model in the Model Registry page.

  2. The New Model page displays.

     

  3. Provide the following details:

  4. Property

    Description

    Model Name

    This field is pre-filled when you are coming from the Model Viewer Page. NOTE: If you set the Model Name to be the same as a model that is already in the Model Registry, then it will be registered as a new version of that model instead of as a new model.

    Model Type

    Select one of the following:

    • “SKLEARN“ for scikit-learn models

    • “KSTUDIO“ for Knowledge Studio models

    • “PYTHON“ for Python-code models.

    The displayed type is that of the latest version of the model

    Model Package Version

    The version of “Sklearn“ or “KStudio“ or “Python“ that the model should be tagged with. This will effect the Docker container that gets deployed on Seldon Core

    Model Repository

    The MLflow internal connection under which the model will be registered

    Model File Location

    Where the user must upload a .pkl file for “Sklearn” and a .py file for “Knowledge Studio“ and “Python“ models

     

  5. Click Save when you are finished.

  6. The new model is registered and displays in the Model Registry.

Adding a New Model Version

Steps:

  1. Click the Add new model version in the Model Viewer page to register a new version for your model.

  2. The Add New Model Version page displays.

     

  3. Use this page to update the:

    • Model type - Sklearn, KStudio, Python

    • Model package version

    • Model file location - Click Browse to open the file explorer, which you can use to select the new model file. Click Open to close this browser and add the selected file to the page.

     

  4. Save your settings and refresh the Model Registry. The updated version number of your model displays.

     

  5. Reselecting the same model to view it shows the older and updated versions of this model in the Model Viewer page.

  6.