feat: add n_input and n_output as keyword arguments

This is needed for compatibility with `@mtkmodel`

Author: SebastianM-C

src/ModelingToolkitNeuralNets.jl

@@ -13,7 +13,7 @@ export NeuralNetworkBlock, multi_layer_feed_forward
 include("utils.jl")
 
 """
-    NeuralNetworkBlock(n_input = 1, n_output = 1;
+    NeuralNetworkBlock(; n_input = 1, n_output = 1,
         chain = multi_layer_feed_forward(n_input, n_output),
         rng = Xoshiro(0),
         init_params = Lux.initialparameters(rng, chain),
@@ -22,8 +22,7 @@ include("utils.jl")
 
 Create an `ODESystem` with a neural network inside.
 """
-function NeuralNetworkBlock(n_input = 1,
-        n_output = 1;
+function NeuralNetworkBlock(; n_input = 1, n_output = 1,
         chain = multi_layer_feed_forward(n_input, n_output),
         rng = Xoshiro(0),
         init_params = Lux.initialparameters(rng, chain),
@@ -46,6 +45,12 @@ function NeuralNetworkBlock(n_input = 1,
     return ude_comp
 end
 
+# added to avoid a breaking change from moving n_input & n_output in kwargs
+# https://github.com/SciML/ModelingToolkitNeuralNets.jl/issues/32
+function NeuralNetworkBlock(n_input, n_output = 1; kwargs...)
+    NeuralNetworkBlock(; n_input, n_output, kwargs...)
+end
+
 function lazyconvert(T, x::Symbolics.Arr)
     Symbolics.array_term(convert, T, x, size = size(x))
 end