lablup / lablup/backend.ai

Model vfolder indexing and browsing

Open
#1,066 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Python
Stars
670
Forks
183
Avg merge
17h 7m
Merged PRs (30d)
358

Description

We need to make the `model-definition.yaml` to be searchable and queriable at large scale, via database queries (GraphQL & REST).

Currently they are stored as plain-text files inside model vfolders, and it is difficult to retrieve their contents in a scalable way from the frontend.

It is impractical to have an automatic bidirectional sync between the file and the database record due to performance and scalability reasons.

Let’s define the synchronization scheme as follows:

- Importing a model vfolder (from model registry) or creating a new one by uploading the file
- Sync: `model-definition.yaml` → database record (model vfolder)
- Modifying the model definition in the Web UI / Exporting a model vfolder (to model registry or as an archive file)
- Sync: database record → overwrite `model-definition.yaml`
- Any change in the vfolder is LOST.

This will be the base implementation for adding customization of model definition per deployment (e.g., differentiated health check conditions, etc.), as well as deployment configuration versioning.

JIRA Issue: BA-220

Contributor guide

Open the contributing guide

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.