WML for z/OS provides extensive built-in
machine learning frameworks to meet the need of the most common data transformation, model training,
and model scoring tasks. If you import models that use additional transformers, estimators, or
algorithms, you can create personalized libraries and runtimes required by those transformers and
estimators.
Procedure
In WML for z/OS, a user-defined library is a
package that provides essential functions for training and scoring a model while a custom-built
runtime consists of a set of libraries. A library package can be a jar or configuration file, such
as a JSON configuration file.
-
Sign into the WML for z/OS web user interface with
your user name and password.
-
From the sidebar, navigate to the
Model Management
page.
-
Create a new library.
-
On the
Model Management
page, click the Runtimes
tab.
-
Click Add Library to launch the setting page for a new library.
-
Specify attributes for the new library and select a library file for upload.
-
Click Create to create the new library.
-
If needed, repeat the procedure to create another new library.
-
Create a new runtime that consists of multiple custom-built libraries.
-
On the
Model Management
page, click the Runtimes
tab.
-
Click New Runtime to launch the setting page for a new runtime.
-
Specify attributes for the new runtime and select your own libraries to be included.
-
Click Create to create the new runtime.
-
If needed, repeat the procedure to create another new runtime.
-
Import a model that will use the new runtime.
-
Select a runtime to be associated with the imported model.
-
Train, deploy, and score the model.
After you successfully deploy the imported model, you can treat and use the model as a normal
WML for z/OS model. You can schedule it for online
scoring, batch scoring, evaluation, and automatic retraining.