Build MLOps Workflows – An improved user experience with Model Manager
Model Manager is the MLOps solution as part of Viya, the AI and Analytics enterprise platform launched in 2017 by SAS.
The objective of an MLOps team is to automate the deployment of ML models into the core software system or as a service component. This means, to automate the complete ML-workflow steps without any manual intervention. Triggers for automated model training and deployment can be calendar events, messaging, monitoring events, as well as changes on data, model training code, and application code.
The SAS MLOps team has worked hard in the past months to offer an improved user experience when building MLOps workflows, adding the latest Workflow Manager as part of Model Manager, the MlOps tool.
Check out this article to know what is Workflow Manager and how it works.
In today's blog, I want to share some of the newest additions over the past months.
In 2021.2.4, we’ve added autosave capabilities, so that your workflow automatically saves upon closing. Also in 2021.2.4, we’ve added the ability to configure custom notification templates. Notification templates define email and message templates that are sent to users when a Send Notification or Send Group Notification task is added to the workflow. This allows workflow designers to populate the notification with actional information that can be used directly for decision making. Workflow designers can embed data objects within the notification, preview the notification, and send a test notification, so that they can be certain that information is flowing into the notification correctly.
In 2021.2.3, we’ve added the ability to include custom attributes within prompts. This allows solutions that use SAS Workflow Manager to pull specific data from the solution into the prompts. For example, SAS Model Manager has two available attributes: show_model_list, which shows a list of models in the current project and select_potential_owner, which launches the Choose Members window. Learn more about these attributes and how to leverage them in the demo video below!
We’ve also enhanced the Instance category view so that Workflow Admins can see when the tasks started and how long they took to complete. This provides insight into bottlenecks within the flow.
And finally, we have been improving how errors are viewed within the Instance category view. Whenever a Service Task fails, the failure is noted on the diagram and table. Clicking the failure icon in the table opens an error pane with the returned error. This helps Workflow Designers better pinpoint why errors are occurring to quickly address and fix them.
What else would you like to see in Workflow Manager? Let me know below!
What to learn more about Workflow Manager? Check out the following links!
