Hello Liquid AI team,

I'm performing some tests on different architectures and wirings of the LNN. However, when I try to use AutoNCP with CfC class, I receive the following error after fitting the model:

Exception has occurred: AttributeError
'WiredCfCCell' object has no attribute 'hidden_units'

The current model only considers a CfC layer and is constructed with AutoNCP as

model = keras.models.Sequential(
    [
        keras.layers.InputLayer(input_shape=input_shape),
        CfC(AutoNCP(units=LNN_SIZE,output_size=output_size,return_sequences=True))
    ])

The dataset has 80000 samples with 20 symbols and 104 features. I'm considering:
input_shape = (20, 104)
output_size = 104

As per my understanding from the previous questions in this repo, CfC layer can work with input shape of this dimension. Is this still true when using AutoNCP?

Many thanks in advance.