Extracting dynamic routing parameters of trained model

[canyalniz]

First of all I would like to thank you for creating such a great resource. I am fairly new to machine learning so sorry if this is an obvious question.

I am trying to extract parameters from a trained model using this code:

weights = []
config = []
for layer in model.layers:
    weights.append(layer.get_weights())
    config.append(layer.get_config())

Conv1_weights = weights[1][0]
Conv1_biases = weights[1][1]

PrimaryCaps_weights = weights[2][0]
PrimaryCaps_biases = weights[2][1]

DigitCaps_weights = weights[5][0]

My question is, how can I get the b and c values used during dynamic routing?

[canyalniz]

I now realize that initially I falsely thought the coupling coefficients were determined during training as well. However I realized that even during inference the model goes through the dynamic routing process to rediscover the coupling coefficients for each new sample.

This is discussed in issue #28 as well.