REGISTER_MODEL

Registers a trained model and adds it to Model Versioning environment with a status of 'under_review'.

Registers a trained model and adds it to Model versioning environment with a status of 'under_review'. The model must be registered by the owner of the model, dbadmin, or MLSUPERVISOR.

After a model is registered, the model owner is automatically changed to Superuser and the previous owner is given USAGE privileges. Users with the MLSUPERVISOR role or dbamin can call the CHANGE_MODEL_STATUS function to alter the status of registered models.

This is a meta-function. You must call meta-functions in a top-level SELECT statement.

Behavior type

Stable

Syntax

REGISTER_MODEL( 'model_name', 'registered_name' )

Arguments

model_name
Identifies the model to register. If the model has already been registered, the function throws an error.
registered_name
Identifies an abstract name to which the model is registered. This registered_name can represent a group of models for a higher-level application, where each model in the group has a unique version number.

If a model is the first to be registered to a given registered_name, the model is assigned a registered_version of one. Otherwise, newly registered models are assigned an incremented registered_version of n + 1, where n is the number of models already registered to the given registered_name. Each registered model can be uniquely identified by the combination of registered_name and registered_version.

Privileges

Non-superusers: model owner

Examples

In the following example, the model log_reg_bfgs is registered to the logistic_reg_app application:

=> SELECT REGISTER_MODEL('log_reg_bfgs', 'logistic_reg_app');
                          REGISTER_MODEL
----------------------------------------------------------------------
Model [log_reg_bfgs] is registered as [logistic_reg_app], version [1]
(1 row)

You can query the REGISTERED_MODELS system table to view details about the newly registered model:

=> SELECT * FROM REGISTERED_MODELS;
  registered_name | registered_version |    status    |        registered_time        |      model_id     | schema_name |    model_name     |      model_type       |    category
------------------+--------------------+--------------+-------------------------------+-------------------+-------------+-------------------+-----------------------+----------------
 logistic_reg_app |                  1 | UNDER_REVIEW | 2023-01-22 09:49:25.990626-02 | 45035996273853740 | public      | log_reg_bfgs      | LOGISTIC_REGRESSION   | VERTICA_MODELS
(1 row)

See also