How to train ML models with Google Vertex AI Training 

Thanks Mahdi

Thanks Vertex AI Pipelines is next

The first Vertex AI Pipeline video is online ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-gtVHw5YCRhE.html

Hi the custom container run under custom jobs within Vertex AI Training. Its not a training pipeline.

Exactly you only pay as long as the job is running. Vertex AI takes care of that. In comparison with a VM you have to take care of that yourself e.g. shutting it down.

Google Cloud Workflows does not support out-of-the-box Vertex AI Training. But you can always invoke a Cloud Run service from within Cloud Workflows and initiate the training job from there. Here is the full list of supported services cloud.google.com/application-integration/docs/all-triggers-tasks#tasks-for-google-cloud-services

Is it optimal or good practice to initiate training job from cloud run? If so, can you add short description on how you'd do it?@@ml-engineer

It's one of possible ways. Just use the vertex AI SDK. Alternative you can use vertex AI Pipelines I have a article and video for that as well

You can test your ML code here is an excellent example medium.com/@burnmg/software-testing-in-tensorflow-2-0-33c440ca908c Vertex AI itself you might can test with an integration test but I do not see any big benefit of it.

Do we have a reference for how to apply unit test for vertex ai pipeline

@@ml-engineer thanks for the reference