In T358676, the Structured Content team completed training a logo detection model and requested the ML team to deploy it on LiftWing. In P58917, both teams collaborated to develop a logo detection prototype, which has been confirmed to meet the Structured Content team's specifications. In this ticket we shall integrate this prototype into KServe and create a model server to be hosted on LiftWing.
Description
Description
Details
Details
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| logo-detection: add KServe custom model-server | machinelearning/liftwing/inference-services | main | +260 -0 |
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Open | None | T349641 [EPIC] MVP Logo machine detection on Commons | |||
| Resolved | kevinbazira | T358676 Host a logo detection model for Commons images | |||
| Resolved | kevinbazira | T361803 Create logo-detection model-server to be hosted on LiftWing |
Event Timeline
Comment Actions
Change #1017453 had a related patch set uploaded (by Kevin Bazira; author: Kevin Bazira):
[machinelearning/liftwing/inference-services@main] logo-detection: add KServe custom model-server
Comment Actions
Change #1017453 merged by Kevin Bazira:
[machinelearning/liftwing/inference-services@main] logo-detection: add KServe custom model-server
Comment Actions
A custom KServe model-server was created for the logo-detection isvc and can be run locally using instructions in this documentation. To keep consistent with other model-servers, the batch input and output are different from what we had in the prototype phase as shown below:
| I/O | Prototype | KServe model-server |
| input | [{img1},{img2},{img3}] | {"instances":[{img1},{img2},{img3}]} |
| output | {[{img1},{img2},{img3}]} | {"predictions":[{img1},{img2},{img3}]} |
In T362598 we are going to containerize this model-server in order to be able to host it on LiftWing.