*   Multiple C++ API updates.
*   Multiple TensorBoard updates including:
    *   Users can now view image summaries at various sampled steps (instead of
        just the last step).
    *   Bugs involving switching runs as well as the image dashboard are fixed.
    *   Removed data download links from TensorBoard.

    Arg<TF_Int64Tensor, [{corresponds to start in python's xrange().}]>:$start,
    Arg<TF_Int64Tensor, [{corresponds to stop in python's xrange().}]>:$stop,
    Arg<TF_Int64Tensor, [{corresponds to step in python's xrange().}]>:$step,

    ConfinedAttr<TypeArrayAttr, [ArrayMinCount<1>]>:$output_types,
    ConfinedAttr<TF_ShapeAttrArray, [ArrayMinCount<1>]>:$output_shapes,
    DefaultValuedOptionalAttr<StrAttr, "\"\"">:$metadata,

{\n\t\t\t\"kind\":\"PluginA\",\n\t\t\t\"aOption\":\"foo\",\n\t\t},\n\t}\n\nSo what happens? Decode first uses json or yaml to unmarshal the serialized data into your external MyAPIObject. That causes the raw JSON to be stored, but not unpacked. The next step is to copy (using pkg/conversion) into the internal struct. The runtime package's DefaultScheme has conversion functions installed which will unpack the JSON stored in RawExtension, turning it into the correct object type, and storing it in the Object....

{\n\t\t\t\"kind\":\"PluginA\",\n\t\t\t\"aOption\":\"foo\",\n\t\t},\n\t}\n\nSo what happens? Decode first uses json or yaml to unmarshal the serialized data into your external MyAPIObject. That causes the raw JSON to be stored, but not unpacked. The next step is to copy (using pkg/conversion) into the internal struct. The runtime package's DefaultScheme has conversion functions installed which will unpack the JSON stored in RawExtension, turning it into the correct object type, and storing it in the Object....