Lab 3: Deploy a Jupyter Notebook
At this point, the cloud platform engineering team has built a self-service mechanism to provision secure environments to host data science projects. The project administrator has provisioned resources using the self-service mechanism for your team to work, and they have provided you with a self-service mechanism to enable you to provision SageMaker notebooks. Now, use these resources to provision a Jupyter notebook and start developing your ML solution.
Launch the notebook product
Navigate to the AWS Service Catalog console and on go to the detail page for the recently provisioned data science environment. Use the hyperlink labeled AssumeProjectUserRole under Outputs to assume the role of a data science user. Assume the role and visit the Service Catalog product listing. Using the notebook product defined for you by the project administrator, launch a Notebook product using the same project name that was used to create the environment.
- Open the Service Catalog provisioned products list.
- Click on the provisioned data science project and locate the hyperlinks in the Output section of the provisioned product detail.
- Click on the AssumeProjectUserRole hyperlink.
- Click
Assume Roleon the next screen. - Return to the Service Catalog product listing.
- Open the menu for the SageMaker notebook product and click
Launch Product - Give the notebook product a name such as
YOUR-NAME-ds-notebookand clickNext - Use the same project name as earlier in the labs
- Enter an email address for
NotebookOwnerEmailand a username forNotebookOwnerUsername - Click
Nextand on the next 2 screens clickNext - On the
Reviewscreen clickLaunch
You will land at a Provisioned Product page while the Service Catalog creates your Jupyter Notebook. Periodically click Refresh until the Status reads Succeeded. This should take about 5 minutes to launch your notebook.
Access the notebook
After the notebook has launched successfully you can open it by clicking the NotebookUrl hyperlink in the Outputs section of the provisoined notebook details page. With your Jupyter notebook open, familiarize yourself with the web interface and open the Notebook kernel named 00_SageMaker-SysOps-Workflow. Don’t forget to reference back to the Jupyter cheat sheet for a quick reference.
When its open the Notebook kernel should use the conda_python3 kernel. If Jupyter asks you to set the kernel select conda_python3 and if Jupyter displays an error, reload the Jupyter page by clicking your browser Refresh button.
Execute the steps in each of the cells of the notebook carefully, reading the text as you work through these notebooks until you get to ## Lab 4: Train Without VPC Configured.
Once that is complete, finish the rest of this notebook.
Takeaways
Having completed the steps in both notebooks, you have learned how to build, train, deploy, and monitor models using Amazon SageMaker with security best practices.