OpenScale Manual Config Part 5
Last updated
Last updated
For a deployed machine learning model, OpenScale will record all of the requests for scoring and the results in the datamart using feedback logging. In this submodule, we'll emulate a production system that has been used for a week to score many requests, allowing the various configured monitors to present some interesting data. Note that this Historic Data submodule can be run at any time.
The submodule contains the following steps:
If you Created the Project using the Customer-Churn-Project.zip file, your notebook will be present in that project, under the Assets tab:
You may now skip to the next step Update credentials
NOTE: You should probably not need this step, and should only perform it if instructed to.
If, for some reason, you are not using the Created the Project step in the Pre-work to import Customer-Churn-Project.zip, then you will need to import the notebook file by itself. Use the following steps for that.
At the project overview click the New Asset button, and choose Add notebook.
On the next panel select the From URL tab, give your notebook a name, provide the following URL, and choose the Python 3.6 environment:
The notebook is hosted in the same repo as the workshop
Notebook: openscale-historic-data.ipynb
Notebook with output: openscale-historic-data-with-output.ipynb
When the Jupyter notebook is loaded and the kernel is ready then we can start executing cells.
In the notebook section 2.0 you will add your Cloud Pak for Data platform credentials for the WOS_CREDENTIALS.
For the url field, change https://w.x.y.z to use the URL your ICP cluster, i.e something like: "url": "https://zen-cpd-zen.cp4d-v5-2bef101da9502000c44fc2b2-0001.us-south.containers.appdomain.cloud".
For the username, use your Cloud Pak for Data login username.
For the password, user your Cloud Pak for Data login password.
Important: Make sure that you stop the kernel of your notebook(s) when you are done, in order to prevent leaking of memory resources!
Spend an minute looking through the sections of the notebook to get an overview. You will run cells individually by highlighting each cell, then either click the Run button at the top of the notebook. While the cell is running, an asterisk ([*]) will show up to the left of the cell. When that cell has finished executing a sequential number will show up (i.e. [17]).
Now that we've simulated a Machine Learning deployment in production, we can look at the associated monitors again and see more detail. Re-visit the various monitors and look again at the graphs, charts and explanations after the addition of the historical data:
With the addition of historical data, we can now use the OpenScale tools in a simulated production environment. We can look at Fairness, Explainability, Quality, and Drift, and see how all transactions are logged. This workshop contains API code, configuration tools, and details around using the UI tool to enable a user to monitor production machine learning environments.
