(tf2/tf.keras) How could # params decreases when a model contains another model

Question

I have a keras Model(not Layer) named Model A. Model A contains keras.Layers of types: Dense, Conv2D, AveragePooling2D, BatchNormalization, add, GlobalAveragePooling2D.

The output of model.summary() is as follows:

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
layer_type_1 (Layer_type_1)  multiple                  3776      
_________________________________________________________________
...                           ...                      ...
_________________________________________________________________
dense (Dense)                multiple                  1024      
=================================================================
Total params: 4,787,808
Trainable params: 4,782,496
Non-trainable params: 5,312
_________________________________________________________________

I have another keras model (model B) which contains model A.

summary():

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d (Conv2D)              multiple                  35        
_________________________________________________________________
model_A (ModelA)             multiple                  4787232   
=================================================================
Total params: 4,787,267
Trainable params: 4,781,955
Non-trainable params: 5,312
_________________________________________________________________

I wonder how could the number of total parameters in model B is less than model A?

Since model B contains model A, it must be greater.

Answer 1

我发现这是因为输入形状（到模型 A）由于模型 B 中的第一个 Conv2D 层而从 (W, D, 6) 更改为 (W, D, 5)。