tf.keras.layers.Conv2D with and without tfpl.IndependentNormal | request for support

I was trying to create a model with and without 'tfpl.IndependentNormal'.

tf.version is : 2.5.0 tfp.version is : 0.13.0

The codes are shared below.
_`def test_model(num_values, feature_size=256, name='test_submodel', tfpl_indep=False):
inputs = tf.keras.layers.Input(shape=(None, None, feature_size))
outputs = inputs
if tfpl_indep:
num_values=tfpl.IndependentNormal.params_size(num_values)
options = {
'kernel_size' : 3,
'strides' : 1,
'padding' : 'same',
'kernel_initializer' : tf.keras.initializers.RandomNormal(mean=0.0, stddev=0.01, seed=None),
'bias_initializer' : 'zeros'
}
outputs = tf.keras.layers.Conv2D( num_values, **options)(outputs)
if tfpl_indep:
outputs =tfpl.IndependentNormal(num_values)(outputs)
test_model=tf.keras.models.Model(inputs=inputs, outputs=outputs, name=name)
return test_model

#calling with and without tfpl.IndependentNormal
reg_model=test_model(num_values=4, tfpl_indep=False)
reg_model_tfpl_indep=test_model(num_values=4, tfpl_indep=True)

input=tf.keras.layers.Input(shape=( 90, 160, 256))

#passing the input to the model
reg_model_input=reg_model(input)
reg_model_input_data
#Works fine with output : KerasTensor(type_spec=TensorSpec(shape=(None, 90, 160, 4), dtype=tf.float32, name=None), name='test_submodel/conv2d_3/BiasAdd:0', description="created by layer 'test_submodel'")

reg_model_tfpl_indep_input=reg_model_tfpl_indep(input)
reg_model_tfpl_indep_input

#this does not give any response and not able to proceed_
`
It would be great if you can guide me to correct this for the tfpl model.

The issue names no repository file or test. Start by reproducing the supplied test_model example with TensorFlow 2.5.0 and TensorFlow Probability 0.13.0, then inspect the tfpl.IndependentNormal.params_size and Keras model call path; done means identifying the compatibility or usage problem and documenting a minimal supported correction.