Workbench instance container

[LinearParadox]

Hi all,

Is it possible to know which container google uses to create it's workbench instances? It seems the one they provide for custom containers is different than the default one. Mainly it seems it uses conda instead of micromamba, and I am logged in as jupyter by default instead of root. It would be helpful to build on top of the one that is used by default.

[ibaui]

Hi @LinearParadox,

Welcome to Google Cloud Community!

Google’s Vertex AI Workbench instances use a Google-provided base container as the default environment which you can start with and modify for your needs. The container is configured with a Container-Optimized OS and includes pre-installed data science packages which might explain the differences you are noticing such as the use of conda for package management instead of micromamba and the default user being jupyter instead of root.

If you want to build on top of this default container, you can derive your container from the provided base image and then build and push it to your Artifact Registry. This approach ensures compatibility with the vertex AI Workbench’s features and services. Here are some helpful documents you can visit to guide you through customizing a container for Google Vertex AI Workbench:

• Creating a Vertex AI Workbench Instance Using a Custom Container: Provides detailed instructions on how to create a custom container derived from a Google-provided base container.
• Introduction to Vertex AI Workbench Instances: Gives an overview of Vertex AI Workbench instances and their capabilities.
• Choosing a Notebook Solution: Discusses different notebook solutions available in Vertex AI Workbench, including custom containers.

I hope the above information is helpful.