Install a model

To enable the installation of an ML model, the admin user must log into the Analytical Workspace. Here you will find a link in the top right-hand corner called 'Advanced Analytics Models library'. To access this library, simply click on the Advanced Analytics Models library link. A page pops up showing the list of all available models. On the left side, you’ll find the list, with a (plus) button at the bottom to add a model. In the central area, you can see the model's description and version.

The following is indicated for each model: the code, whether or not it has the training step, and the status of the model.

The model status can be READY, PENDING, DELETING, DEPLOYED, FAILED, DEPLOY_FAILED, DELETED and OFFLINE.

If, when opening the model library, a model is no longer present in the database, it is shown with the status NOT FOUND.

ATTENTION: The PENDING status is displayed in the red box but does not mean that the model cannot be used for data processing. A newly loaded model has its status set to PENDING and then switches to READY.

To add a model, the admin user must click the '+' button. A popup will then appear, containing a field for the code of the model and one for its description.

To install a new model, click the combobox and all templates that can be installed will be shown. Once the user selects a model, the 'OK' button is enabled. When you click OK, the system installs the model.

If the model is used in an AA, it cannot be deleted.

The ID of the uploaded model can be displayed by clicking on next to the desired model.

When the admin selects the Edit option, a window appears displaying the model ID.

Only the field containing the description of the model is editable.

Model update

The admin can update the model their library with a new version.

To update a model, the following actions must be performed:

1. click on of the model
2. click on Upgrade. The system displays a warning: when clicking OK to confirm, the user may lose the training performed on the model and thus may need to repeat the operation with the updated model.

After selecting the model of interest, its version can be displayed in the central screen under the Model Version section.

IMPORTANT: In case a later version exists, after clicking OK the system starts the update. In case there is no later version than the one currently in the library, the system displays the message: 'There are no new versions for this model'.

If you refresh the page you will see the model in the Upgrading status.

When the page update operation is complete, the model will be in the Ready status and the updated version will be visible in Model Version.

Note: If the operation is unsuccessful, the model remains Ready and at the version before the operation and in the Model Version paragraph the system indicates that the upgrade was unsuccessful.

Locking and unlocking user models

The admin can lock and unlock one or more models in their library.

To lock a model, the following actions must be performed:

1. select the model and click on
2. click on Lock Model The system shows a warning: by clicking on OK, the user will no longer be able to launch Train and Execute processes related to that model and the model will be in the BLOCKED_BY_USER status

Should the user try to launch a data processing related to this model, the data processing will not be executed, because the model's status is different from ready.

To unlock a model, the following actions must be performed:

1. select the locked model and click on
2. click Unlock Model.

Note: This feature is only available for models that are in the Ready status.

Lock from CCH TagetikTagetik

Models can also be locked from CCH Tagetik. In this case, the models will be shown in the BLOCKED_BY_TAGETIK status.

Should the user try to launch a data processing related to this model, the data processing will not be executed, because the model's status is different from ready.

IMPORTANT: The user cannot unlock models that are in the BLOCKED_BY_TAGETIK status.