Continuous-Adaptation-with-VertexAI-AutoML-Pipeline

Continuous-Adaptation-with-VertexAI-AutoML-Pipeline copied to clipboard

Thank you for this great example!

I was wondering what the advantages are of triggering a Kubeflow pipeline over calling Vertex AI endpoints from the cloud function directly? Or is the latter not possible?

@doender

Vertex AI pipeline can be triggered in any ways, and you can benefit many things by triggering it in Cloud Function. For example:

• Cloud Function can be triggered from events from GCP's other services.
• When it is set to be triggered by GCS bucket, whenever there are changes in a bucket, you can decide to trigger the pipeline by putting some conditional logics in the Cloud Function.
• When it is set to be triggered by Cloud Scheduler, you can periodically check if the performance of the currently deployed model degraded, and you can trigger the pipeline based on the threshold you set
• ....