TypeError: 'Tensor' object is not callable | Keras Autoencoder

Question

I'm creating an autoencoder in Keras following this tutorial and I keep getting the following error.

decoder = tf.keras.Model(encoded_input, decoded(input_img))
TypeError: 'Tensor' object is not callable

I believe it's something to do with not being able to use tensors in this way because of the nature of this type of object but I have some gaps in understanding of why and how to go about solving this.

Here is a minimal working example of my code:

# input_img input placeholder
input_img = tf.keras.layers.Input(shape=(16, 16, 1), name ="input")
encoded = tf.keras.layers.Dense(encoding_dim, activation='relu')(input_img)
decoded = tf.keras.layers.Dense(256, activation='sigmoid')(encoded)
autoencoder = tf.keras.Model(input_img, decoded)
encoder = tf.keras.Model(input_img, encoded)
encoded_input = tf.keras.layers.Input(shape=(encoding_dim,))
decoder = tf.keras.Model(input_img, decoded(encoded_input))

OverLordGoldDragon · Accepted Answer · 2019-08-22 18:11:41Z

1

Keras Model expects input & output arguments as layers, not tensors. The tutorial is correct - you appear to have missed a line right before decoder =: -- decoder_layer = autoencoder.layers[-1]

All together, with decoder = changed appropriately:

input_img = tf.keras.layers.Input(shape=(16, 16, 256), name ="input")
encoded   = tf.keras.layers.Dense(encoding_dim, activation='relu')(input_img)
decoded   = tf.keras.layers.Dense(256, activation='sigmoid')(encoded)

autoencoder = tf.keras.Model(input_img, decoded)
encoder     = tf.keras.Model(input_img, encoded)

encoded_input = tf.keras.layers.Input(shape=(encoding_dim,))
decoder_layer = autoencoder.layers[-1] # gets 'decoded' = last layer of 'autoencoder'

decoder = tf.keras.Model(encoded_input, decoder_layer(encoded_input))

edited Aug 22, 2019 at 18:11

answered Aug 22, 2019 at 17:48

OverLordGoldDragon

20k12 gold badges60 silver badges120 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

iiooii Over a year ago

Thanks for the explanation. I'm getting this error now,

ValueError: A target array with shape (1000, 16, 16, 1) was passed for an output of shape (None, 16, 16, 256) while using as loss binary_crossentropy This loss expects targets to have the same shape as the output.

I assume I need to reshape my tensors at some point or change the model architecture so the input and output are identical.

OverLordGoldDragon Over a year ago

Autoencoders seek to reconstruct their input - thus, output & input dims must match. Dense(256) w/ 'sigmoid' yields 256 output predictions - whereas your Input has 1 input label, i.e. last dim in shape=. Fix: Input(shape=(16,16,256)).

Collectives™ on Stack Overflow

TypeError: 'Tensor' object is not callable | Keras Autoencoder

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related