Split layers without Lambda

[mercy0387]

I build the model which has split and concat in itself. This issue ( #890 ) is very helpful for me but there's a problem.

I want to use my model in iOS App. So I tried to convert keras model to coreML model using coremltools.
But coreML does not accept Lambda layers like this ...

ValueError: Keras layer '<class 'keras.layers.core.Lambda'>' not supported. 

I just want to do

input1 = Lambda(lambda x: x[:, :5], output_shape=(5,))(one_input)
input2 = Lambda(lambda x: x[:, 5:], output_shape=(5,))(one_input)

and after some manipulation

output = Concatenate()([layer1, layer2])

Is there work around for this without using Lambda?

[mercy0387]

I think I can say the same thing to backend functions.

layer = Lambda(lambda x: K.sqrt(x))(input_layer)

Can I use alternative layers like these in coreML?