Challenge 1: Let’s start exploring!
Introduction
As depicted in the overview diagram, the first step of any ML project is data analysis and maybe some experimentation. Jupyter notebooks are great for interactive exploration. We can run those locally, but Vertex AI provides managed environments where you get to run Jupyter with the right security controls and flexible hardware options.
Description
Create a Vertex AI Workbench Instance. Pick a region close to you and choose the single user only option.
It’s a good practice to have isolated virtual environments for experiments, so create a new virtual environment and install that as a kernel. See this gist for the instructions.
Warning Not using a dedicated and isolated environment/kernel might cause dependency conflicts as Vertex AI Workbench Instances come pre-installed with some versions of the required libraries.
We’ve prepared a sample project on Github, navigate there and download the project as a zip file and extract the contents of the zip file onto your Notebook instance. Open the notebook 01-tip-toe-vertex-ai.ipynb
, make sure that you’ve selected the newly created kernel. You should now be able to run the first notebook and get familiar with some of the Vertex AI concepts.
Note As we’re installing packages in the first cell of the sample notebook and restarting the kernel in one of the following cells, Run All will not work. Run the cells one-by-one also to understand what’s going on in every cell.
Success Criteria
- There’s a new single-user Vertex AI Workbench Instance.
- The sample notebook
01-tip-toe-vertex-ai.ipynb
is successfully run (using the newly generated kernel) and a model file is generated/stored in Google Cloud Storage. - No code was modified.
Tips
- Some of the required settings can be found in the Advanced Settings section when you’re creating a new Vertex AI Workbench Instance.
- If there’s nothing mentioned in the instructions about a parameter, stick to the defaults (this applies to all of the challenges).
- You can download the zip file to your local machine and then upload it to the Notebook instance, but you can also get the zip URL and use the
wget
(orcurl
) command from a terminal on the Notebook instance. - The sample notebook creates a bucket in a specific region, take note of that as you’ll need that information in the next challenges.
Learning Resources
- Documentation on Vertex AI Workbench Instances