Google Cloud VertexAI

How to execute individual steps in Vertex AI

This is an older version of the ZenML documentation. To read and view the latest version please visit this up-to-date URL.

The Vertex step operator is a step operator flavor provided with the ZenML gcp integration that uses Vertex AI to execute individual steps of ZenML pipelines.

When to use it

You should use the Vertex step operator if:

• one or more steps of your pipeline require computing resources (CPU, GPU, memory) that are not provided by your orchestrator.

• you have access to Vertex AI. If you're using a different cloud provider, take a look at the SageMaker or AzureML step operators.

How to deploy it

• Enable Vertex AI here.

• Create a service account with the right permissions to create Vertex AI jobs (roles/aiplatform.admin) and push to the container registry (roles/storage.admin).

How to use it

To use the Vertex step operator, we need:

• The ZenML gcp integration installed. If you haven't done so, run

  zenml integration install gcp

• Docker installed and running.

• Vertex AI enabled and a service account file. See the deployment section for detailed instructions.

• A GCR container registry as part of our stack.

• (Optional) A machine type that we want to execute our steps on (this defaults to n1-standard-4). See here for a list of available machine types.

• A remote artifact store as part of your stack. This is needed so that both your orchestration environment as well as VertexAI can read and write step artifacts. Check out the documentation page of the artifact store you want to use for more information on how to set that up and configure authentication for it.

• A local orchestrator as part of your stack. This is a current limitation of the Vertex step operator which will be resolved in an upcoming release.

We can then register the step operator and use it in our active stack:

zenml step-operator register <NAME> \
    --flavor=vertex \
    --project=<GCP_PROJECT> \
    --region=<REGION> \
    --service_account_path=<SERVICE_ACCOUNT_PATH> \
#   --machine_type=<MACHINE_TYPE> # optionally specify the type of machine to run on

# Add the step operator to the active stack
zenml stack update -s <NAME>

Once you added the step operator to your active stack, you can use it to execute individual steps of your pipeline by specifying it in the @step decorator as follows:

from zenml.steps import step

@step(custom_step_operator=<NAME>)
def trainer(...) -> ...:
    """Train a model."""
    # This step will be executed in Vertex.

ZenML will build a Docker image called <CONTAINER_REGISTRY_URI>/zenml:<PIPELINE_NAME> which includes your code and use it to run your steps in Vertex AI. Check out this page if you want to learn more about how ZenML builds these images and how you can customize them.

A concrete example of using the Vertex step operator can be found here.

For more information and a full list of configurable attributes of the Vertex step operator, check out the API Docs.