Is there any way to allow alpha channel for CoreML image output?

I'd like my CoreML model image output to allow 'RGBA' color space.
But I couldn't find the way how to do it directly, except converting from MLMultiArray to 'RGBA' colorspace image.

I found that ImageFeatureType colorspace doesn't have 'RGBA' value yet.
https://github.com/apple/coremltools/blob/master/coremltools/proto/FeatureTypes_pb2.py#L27

Is there any way to achieve it?
Or could you add the feature to allow 'RGBA' image type output?

Start by reviewing the ImageFeatureType definition in coremltools/proto/FeatureTypes_pb2.py and the surrounding Core ML image output handling. Determine whether RGBA output can be supported there and identify any validation or conversion paths that constrain image color spaces. Done means CoreML image outputs can represent RGBA without requiring manual MLMultiArray conversion.

Written by the indexing model from the issue text.