ODSC-39737 : allow to use GenericModel.predict() locally

[mingkang111]

Description

Related JIRA:https://jira.oci.oraclecorp.com/browse/ODSC-39737
Currently, the GenericModel supports two methods that can be used to test the model:

• verify()
• predict()

The difference between these two methods is that verify is used for local testing, where predict sends the request to the ML endpoint. Having two different methods confuses users, they don't really understand why we need two different methods for inferring the model.

Changed

• Add a new attribute - local=True/False to the predict method to be able to invoke the prediction locally.

User Experience

prediction = GenericModel.predict(local=True)

Test

Int test in TC

TC: https://teamcity.oci.oraclecorp.com/buildConfiguration/Odsc_Ads_AdsCiCd_CiTrigger/38127161?buildTab=buildParameters

[mayoor]

As a followup, I have a proposal -

If user calls predict and the url is None, when we throw an error, we can provide a message saying, Error invoking the remote endpoint as the model is not deployed yet or the endpoint information is not available. If you intend to invoke inference using locally available model artifact, set parameter local=True

[mingkang111]

As a followup, I have a proposal -

If user calls predict and the url is None, when we throw an error, we can provide a message saying, Error invoking the remote endpoint as the model is not deployed yet or the endpoint information is not available. If you intend to invoke inference using locally available model artifact, set parameter local=True

That's a good idea! I just improved the error message as suggested. Thanks Mayoor.