How do I set the weights from an autoencoder to a CNN model?

Question

So I am trying to implement a model based on the codes found here:

All is well until I get to this point:

for l1, l2 in zip(full_model.layers[:19], autoencoder.layers[0:19]):
  l1.set_weights(l2.get_weights())

And I get this error, even though the full model has the same weights as the autoencoder:

ValueError: You called `set_weights(weights)` on layer "flatten_10" with a weight list of length 2, but the layer was expecting 0 weights. Provided weights: [array([[[[-3.37540284e-02,  7.36398697e-02, -1.45...

The weights for the autoencoder and the full model are as follows:

autoencoder.get_weights()[0][1]

array([[[-0.07640466,  0.08604569,  0.10095344,  0.08002567,
         -0.05795965,  0.12277777,  0.04575707,  0.1368851 ,
          0.08104218,  0.04106109,  0.01466343, -0.00301184,
         -0.02941842, -0.06449406, -0.11245678,  0.03759771,
         -0.04456315,  0.05147151, -0.05671669,  0.03154052,
          0.08367646, -0.03407011,  0.03081554, -0.07092344,
         -0.0342903 , -0.12681712, -0.11921115, -0.00943625,
          0.07913507, -0.11182833,  0.06839333, -0.10381861]],

       [[ 0.12951587, -0.10705423,  0.14214374,  0.10236198,
          0.04869333, -0.07741497,  0.04825569,  0.140887  ,
         -0.04529881,  0.10183885,  0.09898531,  0.0463811 ,
         -0.0497799 , -0.03215659, -0.1106519 ,  0.0191465 ,
         -0.03108089,  0.11891119,  0.13607842, -0.06900101,
          0.02550365, -0.07291926,  0.0408677 , -0.13281997,
         -0.10269159,  0.12453358, -0.06403439,  0.03591786,
          0.09293085,  0.04930058, -0.07233981, -0.11631108]],

       [[ 0.09462225, -0.13031363, -0.07633019,  0.07383946,
         -0.08967619,  0.03298028,  0.05059863, -0.07996925,
         -0.0285711 , -0.02666069, -0.02046945, -0.02898544,
         -0.0632349 ,  0.01124811, -0.06102825, -0.02444353,
         -0.02901937,  0.07315389,  0.04660689, -0.03481405,
          0.03801505, -0.02921393,  0.03578328,  0.00787276,
         -0.13757674, -0.01068925, -0.10495549, -0.04071948,
         -0.01119018,  0.02144167,  0.09804168, -0.05260663]]],
      dtype=float32)

full_model.get_weights()[0][1]

array([[[-0.07640466,  0.08604569,  0.10095344,  0.08002567,
         -0.05795965,  0.12277777,  0.04575707,  0.1368851 ,
          0.08104218,  0.04106109,  0.01466343, -0.00301184,
         -0.02941842, -0.06449406, -0.11245678,  0.03759771,
         -0.04456315,  0.05147151, -0.05671669,  0.03154052,
          0.08367646, -0.03407011,  0.03081554, -0.07092344,
         -0.0342903 , -0.12681712, -0.11921115, -0.00943625,
          0.07913507, -0.11182833,  0.06839333, -0.10381861]],

       [[ 0.12951587, -0.10705423,  0.14214374,  0.10236198,
          0.04869333, -0.07741497,  0.04825569,  0.140887  ,
         -0.04529881,  0.10183885,  0.09898531,  0.0463811 ,
         -0.0497799 , -0.03215659, -0.1106519 ,  0.0191465 ,
         -0.03108089,  0.11891119,  0.13607842, -0.06900101,
          0.02550365, -0.07291926,  0.0408677 , -0.13281997,
         -0.10269159,  0.12453358, -0.06403439,  0.03591786,
          0.09293085,  0.04930058, -0.07233981, -0.11631108]],

       [[ 0.09462225, -0.13031363, -0.07633019,  0.07383946,
         -0.08967619,  0.03298028,  0.05059863, -0.07996925,
         -0.0285711 , -0.02666069, -0.02046945, -0.02898544,
         -0.0632349 ,  0.01124811, -0.06102825, -0.02444353,
         -0.02901937,  0.07315389,  0.04660689, -0.03481405,
          0.03801505, -0.02921393,  0.03578328,  0.00787276,
         -0.13757674, -0.01068925, -0.10495549, -0.04071948,
         -0.01119018,  0.02144167,  0.09804168, -0.05260663]]],
      dtype=float32)

I am trying to tune it according to my dataset and the changes I made with the code are:

#creating the encoder function
def encoder(input_img):
  x = Conv2D(32, (3, 3), activation='relu', padding='same')(input_img)
  x = MaxPooling2D((2, 2))(x)
  x = Conv2D(64, (3, 3), activation='relu', padding ='same')(x)
  x = MaxPooling2D((2, 2))(x)
return x

#creating the decoder function
def decoder(x):
  x = Conv2D(64, (3, 3), activation='relu', padding='same')(x)
  x = UpSampling2D((2,2))(x)
  x = Conv2D(32, (3, 3), activation='relu', padding='same')(x)
  x = UpSampling2D((2,2))(x)
  decoded = Conv2D(1, (3, 3), activation='sigmoid', padding='same')(x)
return decoded

and

autoencoder_train = autoencoder.fit(X_train, X_train,
                                    batch_size = 256,
                                    epochs = 10,
                                    verbose = 1,
                                    validation_data = (X_val, X_val))

where, instead of training the dataset with its labels, I trained it using the same training set. How can I make the full model accept the weights of the autoencoder? I tried training the autoencoder with the input labels, but it does not learn the input.

Answer 1

Nevermind, I know where the problem is. I just needed to change the range of layers in this line,

for l1, l2 in zip(full_model.layers[:19], autoencoder.layers[0:19]):
  l1.set_weights(l2.get_weights())

according to how many layers my encoder has. In this case, it should be full_model.layers[0:5], instead of full_model.layers[0:19]