Layer names and output columns cannot be set as the same. #9
Description
This was an internal decision to make the building of the DAG that builds the keras pipeline simple.
However it is not ideal, and causes us to also add an identity layer to the outputs of the keras pipeline so that the output names match the Spark pipeline.
Instead we can add random strings to the layer name and output name while building the DAG/pipeline, since the user never sees these internals. We then need to remove the identity layer applied to the outputs (since otherwise we can have two layers with the same name), and thus this would be a breaking change.