Way of providing input / output arrays for "Functional Models"

[drout0]

There should be a way to provide the Model with two arrays (inputs, outputs) in order to facilitate (or eliminate) passing of these parameters to Model(), model.compile() and model.fit().
The side-effect is also eliminating the (high) risk of the programmer passing the parameters in the wrong order.
The internals could be something like this:

instead of:

model = Model( inputs = [inputs, aux_inputs ], outputs = [ outputs, aux_outputs ])
model.compile( loss = { 'outputs': 'binary_crossentropy', 'aux_outputs': 'mse' },
               loss_weights = { 'outputs': 1, 'aux_outputs': 0.4 }, metrics = [ 'accuracy' ])
history = model.fit( { 'inputs': X_train, 'aux_iputs': X_aux },
                     { 'outputs': y_train, 'aux_outputs': y_aux },
                     validation_split = 0.2, epochs = 20 )

replace by:

(          io,           ioDataset,  ioLoss,            ioWeights ) = range( 4 )  # cheap way of using enums, but enums would be preferrable
inps =  [[ inputs,       X_train ],
         [ aux_inputs,   X_aux   ]]
outps = [[ outputs,      y_train,    binary_c_entropy,  1   ],
         [ aux_outputs,  y_aux,      mse,               0.4 ]]

model = Model( inputs = inps[ :, io ], outputs = outps[ :, io ] )
model.compile( loss = outps[ :, ioLoss ], loss_weights = outps[ :, ioWeights ], metrics = [ 'accuracy' ])
history = model.fit( [ inps[ :, ioDataset ], outps[ :, ioDataset ], validation_split = 0.2, epochs = 20 )