Skip to content

Latest commit

 

History

History
90 lines (54 loc) · 4.1 KB

odsc-4-deploy-model.md

File metadata and controls

90 lines (54 loc) · 4.1 KB

Lab 4 - Deploy A Model and Invoke It

Introduction

This lab will guide you through deploying a cataloged model from the OCI Console and then invoking it.

Estimated lab time: 15 minutes

Objectives

In this lab you will:

  • Learn how to deploy a model from the console
  • Invoke the model endpoint

Prerequisites

  • You are signed-in to Oracle Cloud
  • You have navigated to Data Science
  • You have opened the Data Science project
  • You have a notebook session in your project
  • You have a model stored in the model catalog

STEP 1: Deploy the model

Now we're going to deploy the model to its own compute instance so that it can be used. This will take the model from the catalog and create a runtime version of it that's ready to receive requests.

  1. Confirm you have completed all the prerequisites and are viewing your Data Science project.

  2. Under Resources, select Models to see the list of models in your model catalog.

  3. Find the row containing sklearn-employee-attrition. On the right end of the row, click the 3-dot icon to open a pop-up menu.

  4. In the pop-up menu, click Create model deployment. The Create Model Deployment dialog opens.

  5. In the Create Model Deployment dialog, configure the fields as described below.

    • Ensure Compartment is set to root.
    • Enter Name as MyModelDeployment
    • Ensure Models is set to sklearn-employee-attrition
    • In the Compute box, click Select.
    • Enter 1 for Number of Instances and check VM.Standard2.4. Then click Submit.

    • (We will not configure Logging in this lab.)

    • Click Create in the Create Model deployment dialog. It takes about 10 minutes for the deployment to be provisioned and its status will be Creating. Wait for the status to change to Active to indicate the model is deployed.

    Note - You just deployed the model through the OCI Console, but you can also deploy a model using python code in a notebook. The ADS library provides functions to do this. See Model Deployment with ADS for more information.

  6. When the model is Active, copy the HTTP endpoint URI to your clipboard (to be used in the next step). To do that look under Resources and click Invoking Your Model. Your model endpoint is displayed. Use the Copy button to copy it to your clipboard.

STEP 2: Invoke the deployed model

Now that the model is deployed, we will go back to the notebook and invoke its HTTP endpoint.

  1. If your notebook session tab in your browser is still open from the previous lab, then view it. Otherwise, follow these steps to return to it.

    • Under Resources, click Notebook Sessions.
    • Click on the notebook session named LabNotebookSession to open it.
    • Click on Open. It will open in a separate browser tab. If prompted to sign-in, provide your Oracle Cloud credentials.

  2. Ensure you are viewing the browser tab/window displaying LabNotebookSession.

  1. Open notebook 2-model-deployment.ipynb, but don't start stepping through it yet.

  2. In the next-to-the-last code block, find the statement:

    # Replace with the uri of your model deployment:
    uri = ''

  3. Paste your deployed model endpoint URI into the statement between the single quote marks.

  4. Execute each code block of the notebook.

  5. Review the output and see that your deployed model returned attrition predictions for the submitted input data.

Congratulations! You have successfully built, trained, cataloged, deployed, and invoked a machine learning model with OCI Data Science.

You may proceed to the next lab.