Getting Started with using IBM Watson OpenScale in under 10 minutes!
IBM Watson OpenScale helps you manage production AI with trust and confidence in outcomes. It tracks and measures outcomes from AI across its lifecycle, and adapts and governs AI to changing business situations — for models built and running anywhere. In this article we will outline how you can get started with using IBM Watson OpenScale in under 10 minutes! So let's get started.
IBM Watson OpenScale is available in public cloud as well as in IBM Cloud Pak for Data which can run behind your companies firewall. In this article we will showcase how you can get started to using OpenScale in IBM Cloud in 4 easy steps.
Step 1: Login to IBM Cloud
Navigate to https://www.ibm.com/in-en/cloud and login using your IBM ID.
Step 2: Find IBM Watson OpenScale in the Catalog and create an instance.
In the top menu bar (see Figure 1) click on Catalog which will show the list of offerings available in IBM Cloud. Search for OpenScale in the catalog (Figure 2) and it will show the Watson OpenScale offering.
Click on the OpenScale tile to see the details of the offering (Figure 3).
There are two plans of OpenScale: The Lite plan and the Standard Plan. The details of each of these plans is available on this page. The Lite plan is free of cost and has some usage limits. This plan is selected by default. Click on the Create button to create a Watson OpenScale service instance. Please note that there can only be a single instance of OpenScale in a given IBM Cloud account.
Step 3: Launch and run Auto Setup
Once the OpenScale instance is created you will be taken to the OpenScale service details page (shown in Figure 4).
Click on the Launch Application to launch the OpenScale GUI. The very first screen will ask if you would like to run Auto Setup.
Click on the Auto Setup. This does multiple things to get OpenScale up and running. Some of the important things that it does under the covers are:
- OpenScale needs a serve engine where models are deployed. OpenScale supports multiple serve engines such as Watson Machine Learning, Azure, Sagemaker, SPSS as well as custom serve engine (which can be used to monitor models deployed on say a spark cluster or anywhere other than the supported serve environments). Auto Setup checks if there already exists a WML instance in the account. If not it creates a free WML instance.
- OpenScale needs a model to be deployed in the serve engine so that it can monitor it. Hence Auto Setup creates and deploys a model in the created WML instance.
- OpenScale needs to know about the serve engine and how to connect to it. Therefore Auto Setup registers the WML instance as a serve engine in OpenScale. It makes use of an inbuilt data mart for storing all model payload data as well as metrics computed by OpenScale. Ideally customers use their own database such as PostgreSQL or Db2 as their data mart.
- Auto Setup then creates a subscription to the model in OpenScale so that OpenScale can start monitoring the model.
- Auto Setup then configures different fairness monitors such as fairness, accuracy, etc., so that OpenScale can start computing metrics for the model.
Once the above steps are complete, OpenScale is ready to use!
Step 4: Take a tour to understand features of OpenScale
Once the Auto Setup is complete you can take a tour of OpenScale which gives an overview of all its features.
Thus with just a few clicks you can setup OpenScale and start using it. You can then register your own models in OpenScale and start monitoring them for Bias, generate explanations, find drifts, find accuracy metrics, etc.
Happy AI monitoring!
