Capability to define output key for Importer

Describe the feature and the current behavior/state. Currently, the output dict of the importer defaults to result. It will be good to let the user specify the output dict.

Will this change the current API? How? Adding a new arg, output_key='result' to the Importer API.

Who will benefit with this feature? This will enable reusing the existing code snippet. e.g. Current behavior requires the output dict to be updated everywhere:

schema_gen = Importer(...)
# schame_gen = SchemaGen(...)

trainer = Trainer(..., schema=schema_gen['result'], ...)
# trainer = Trainer(..., schema=schema_gen['schema_gen'], ...)

evalutor = Evaluator(..., schema=schema_gen['result'], ...)
# evalutor = Evaluator(..., schema=schema_gen['schema_gen'], ...)

Do you have a workaround or are completely blocked by this? : Refactor the code

Name of your Organization (Optional) Twitter, Inc.

[sumitbinnani]

In addition to SchemaGen, also ExampleGen. Reusing a transformed training dataset is also a common use case. Similarly, if one have multiple trained models that they would like to use for evaluation, one might want to also import channel of type Model. etc.

[UsharaniPagadala]

@sumitbinnani

Could you please move this to closed status if it is resolved.Thanks