search

kubeflow / pipelines

Machine Learning Pipelines for Kubeflow
https://www.kubeflow.org/docs/components/pipelines/
Apache License 2.0
3.56k stars 1.61k forks source link

ML-Engine deploy model component missing region parameter #5304

Closed Amrit-Mann closed 2 months ago

Amrit-Mann commented 3 years ago

Hi, I am using the ml-engine component to deploy a ML model to AI Platform Prediction (from a Kubeflow deployment in AI Platform Pipelines), however there is no available parameter to deploy the model into a particular region. This results in the model deployment defaulting to the us-central1 region. Is there any way to deploy a model using this component to a specific region? Any help on this will be much appreciated!. Thanks.

Link to component: https://github.com/kubeflow/pipelines/tree/master/components/gcp/ml_engine/deploy

What steps did you take: 1 - Trained a model using AI Platform Pipelines 2 - Saved this model to Google Cloud Storage as a .pkl file 3 - Tried to use the deploy component to deploy this model to AI Platform Prediction, however What happened: There is no region parameter for this deploy component hence the deployment happens in the default region (us-central1) What did you expect to happen: Be able to specify the region parameter in the input for this component. (Since we can do this when deploying the model from either gcloud commands/GCP Console Anything else you would like to add: /kind feature /kind component

NikeNano commented 3 years ago

/assign

NikeNano commented 3 years ago

@zijianjoy should we aim for making all arguments available for the user or keep a selected few(I see that we currently only support some options)?

NikeNano commented 3 years ago

After looking in to this I believe this is actually already possible @Amrit-Mann, if you look at the options: for model and version the take a json following:

where in you have the options to specify the region that you like to deploy to.

@zijianjoy should we aim for making all arguments available for the user or keep a selected few(I see that we currently only support some options)?

All available options should already be possible to set. Let me know if you still has some issues @Amrit-Mann

stale[bot] commented 3 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

zijianjoy commented 3 years ago

Thanks @NikeNano for the research! Since you can use ML-engine REST API, looks like you should be able to call this endpoint to specify regions.

stale[bot] commented 2 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

github-actions[bot] commented 2 months ago

This issue has been automatically closed because it has not had recent activity. Please comment "/reopen" to reopen it.