Extract learned NN posterior weight distribution parameters from DenseVariational layer(s)

I'm working with a model that has a DenseVariational layer defined as follows (lifted from examples found online):

def posterior_mean_field(kernel_size, bias_size=0, dtype=None):
    n = kernel_size + bias_size
    c = np.log(np.expm1(1.))
    return tf.keras.Sequential([
        tfp.layers.VariableLayer(2 * n, dtype=dtype),
        tfp.layers.DistributionLambda(lambda t: tfd.Independent(
            tfd.Normal(loc=t[..., :n], scale=1e-5 + tf.nn.softplus(c + t[..., n:])),
            reinterpreted_batch_ndims=1)),
    ])


def prior_trainable(kernel_size, bias_size=0, dtype=None):
    n = kernel_size + bias_size
    return tf.keras.Sequential([
        tfp.layers.VariableLayer(n, dtype=dtype),
        tfp.layers.DistributionLambda(lambda t: tfd.Independent(tfd.Normal(loc=t, scale=1),
                                                                reinterpreted_batch_ndims=1)),
    ])

dense = tfp.layers.DenseVariational(units=units, make_posterior_fn=posterior_mean_field,
                                                             make_prior_fn=prior_trainable,
                                                            )(prev_layer)

If I train my model and then remove the layers following this layer, the remaining model will output random variables from the learned posterior weight distributions. Something like this:

from tensorflow.keras import Model
# DenseVariational layer is 3rd to last layer in this case
cropped_model = Model(inputs, model.layers[-3].output)  
cropped_mode.predict(test_data)

Most of the time this is fine (e.g. training, sampling, etc.). However, is there a direct way to get the learned loc and scale posterior values returned for a given input (e.g. test_data) to this cropped_model, instead of a sample draw from the distribution they define?

Start with the DenseVariational layer and the posterior function shown in the issue, then trace how the layer exposes its learned distribution during prediction. Determine whether the requested loc and scale values can be returned for test_data without sampling, and document or test the supported behavior.