GoogleCloudPlatform / vertex-ai-alphafold-inference-pipeline

This repository compiles prescriptive guidance and code samples demonstrating how to operationalize AlphaFold batch inference using Vertex AI Pipelines.
Apache License 2.0
67 stars 30 forks source link

Allow specifying Image URI for Alphafold components through environment variable #44

Open sosoihd opened 9 months ago

sosoihd commented 9 months ago
Capture d’écran 2024-02-16 à 17 35 54

Description:

The current Alphafold Portal on GCP uses a predefined approach to construct the IMAGE_URI for Alphafold components, either from GCR.io or Artifact Registry. However, this limits users who store their Docker images in a different Artifact Registry location, outside the project and potentially even in a different region.

Problem:

In my case, the Docker images are centralized in a dedicated Artifact Registry belonging to a different project. The current logic within the portal restricts me to using either GCR.io or the Artifact Registry within the same project, making it impossible to leverage my preferred image location.

Proposed Solution:

I propose introducing an environment variable named IMAGE_URI that allows users to directly specify the desired image URI for Alphafold components. This would grant us the flexibility to utilize images stored in any accessible Artifact Registry, regardless of project or region boundaries.

Benefits:

Enhanced Flexibility: Users gain the freedom to choose their preferred Artifact Registry location, catering to diverse project structures and image management preferences. Accommodate Diverse Use Cases: This addition caters to scenarios where images are stored outside the project or in a different region, expanding the portal's applicability. Additional Information:

I believe this change would significantly increase the flexibility and cater to a wider range of user workflows within the Alphafold Portal.

Thank you