Chapter 1. Overview of developing a data model

Read this section to understand the work required to develop and deploy an application that uses a predictive model created using Red Hat OpenShift Data Science.

Your organization might split responsibility for this process between several roles, such as a data scientist and an application developer, or this work might be done by a single role. An appropriate role is noted for each step.

Table 1.1. Development tasks by role

Application developerData scientistTask description

 

Create a Python S2I project in Git using an OpenShift Data Science application template.

 

Configure user access to the Git project so that data scientists can push to and pull from the repository.

From this point, you can develop the model and the application that uses it simultaneously.

 

Create an OpenShift application using the project repository.

 

Build the OpenShift application to verify your code.

 

Automate the build process using webhooks.

 

Launch Jupyter and either create or import a notebook.

 

Import the application Git project into JupyterLab.

 

Develop and test your model using notebooks in JupyterLab.

 

Save your model as an independent Python function in a separate Python file.

 

Update the requirements.txt file with dependencies your function requires.

 

Test the function on your notebook server.

 

Push your updates back to the remote Git project.

Test the deployed application endpoint.