Create a workbench

Inside the new project, we can create different components. One of them, and probably the most important one when it comes to model training, will be the Workbench. A workbench is a containerized workspace that operates within a pod and contains a collection of different AI/ML libraries and a JupyterLab server.

  1. There are different tabs inside our project page. Click on Workbenches.

  2. On there, select the Create workbench button to open the configuration window.

    Create Workbench

  3. Complete the fields as follows. You will see that there are plenty of different configurations depending on your model and environment specifications:

    • Name: Type your preferred name. I will use model-training.

    • Image selection: the election will depend on the model requirements. Click on View package information to see the packages included. Our model will use Jupyter | TensorFlow | CUDA | Python 3.12.

    • Version selection: The recommendation is to use the latest package versions. That will be the 2025.2 version.

    • Container size: we can select the resources assigned depending on our hardware: number of CPUs, memory, and request capacity of the container. In this case, the Small size is enough.

    • Make sure a persisten volume called model-training-storage is going to be created with 20GB.

    • Under connections, select Attach existing connections. Verify that the vehicle-models connection is selected and click Attach.

  4. Review your configuration and press Create workbench. This last step will trigger the workbench creation.

  5. You will be automatically redirected to another page where you can track the workbench deployment process. Wait until the Status shows Running and click model-training.

    Workbench status

  6. A new tab will open. Enter again your OpenShift credentials to authenticate to the workbench’s JupyterLab platform.

    • User: {openshift_cluster_admin_username}

    • Password: {openshift_cluster_admin_password}

JupyterLab landing page