`nn_name` argument does not work in `SymbolicNeuralNetwork`

[TorkelE]

I can update the name of the neural network parameter, but not the neural network parameter itself.

# Preparation
using Lux, ModelingToolkitNeuralNets, StableRNGs, ModelingToolkit
rng = StableRNG(123)
chain = Lux.Chain(
    Lux.Dense(1 => 3, Lux.softplus, use_bias = false),
    Lux.Dense(3 => 3, Lux.softplus, use_bias = false),
    Lux.Dense(3 => 1, Lux.sigmoid_fast, use_bias = false)
)

# Default names.
NN, NN_p = SymbolicNeuralNetwork(; chain, n_input = 1, n_output = 1, rng)
ModelingToolkit.getname(NN) # :nn_name
ModelingToolkit.getname(NN_p) # :p

# Trying to set specific names.
nn_name = :custom_nn_name
nn_p_name = :custom_nn_p_name
NN, NN_p = SymbolicNeuralNetwork(; chain, n_input = 1, n_output = 1, rng, nn_name, nn_p_name)
ModelingToolkit.getname(NN) # :nn_name # Should be :custom_nn_name
ModelingToolkit.getname(NN_p) # :custom_nn_p_name
`

[SebastianM-C]

Looking at

julia> @macroexpand @parameters ($(nn_name)::typeof(wrapper))(..)[1:n_output] = wrapper
quote
    var"##nn_name#263" = (ModelingToolkit.toparam)((Symbolics.wrap)((setmetadata)((Symbolics.setdefaultval)((setmetadata)((Symbolics.CallWithMetadata)((Sym){(SymbolicUtils.FnType){Tuple, Array{Real, (length)((1:n_output,))}, typeof(wrapper)}}(nn_name)), Symbolics.ArrayShapeCtx, (1:n_output,)), wrapper), Symbolics.VariableSource, (:parameters, :nn_name))))
    [var"##nn_name#263"]
end

I think this is an upstream bug in the macro. (cc @AayushSabharwal )

The interesting thing is that nn_name is properly used once during that and the show method somehow displays custom_nn_name:

julia> NN
custom_nn_name⋆

[TorkelE]

Any updates on this (@AayushSabharwal , @SebastianM-C)? I was trying to fill in a realted bug report, but this kind of confussed me and I am not sure if this is the actual error, I have a different error, or everything is working as intended.

[AayushSabharwal]

JuliaSymbolics/Symbolics.jl#1673