TF 2.3 Concatenate error - AttributeError: 'Sequential' object has no attribute '_nested_outputs'

Question

fusion_model = tf.keras.layers.concatenate([m1.output, m2.output])

m1 and m2 are Model objects.

I'm concatenating the outputs of the two models and I'm getting the following error:

AttributeError: 'Sequential' object has no attribute '_nested_outputs'

This exact code used to work in TF 1.14. Any help? thanks

After experimenting with this example, it appears to crash on the statement m2.output (m1.output runs fine)

my M2 model is:

Layer (type)                 Output Shape              Param #   
=================================================================
rescaling_1 (Rescaling)      (None, 456, 456, 3)       0         
_________________________________________________________________
efficientnetb5 (Functional)  (None, 15, 15, 2048)      28513527  
_________________________________________________________________
global_average_pooling2d (Gl (None, 2048)              0         
_________________________________________________________________
dense_3 (Dense)              (None, 160)               327840    
_________________________________________________________________
dropout_19 (Dropout)         (None, 160)               0         
_________________________________________________________________
predictions (Dense)          (None, 19)                3059

Do you mind sharing a code snippet so we can reproduce the error? — stephen_mugisha
– stephen_mugisha, Commented Sep 7, 2020 at 20:10
@stephen_mugisha Hi can't provide a full example but this is the best I can provide for now :) — Christo S. Christov
– Christo S. Christov, Commented Sep 7, 2020 at 21:03

Christo S. Christov · Accepted Answer · 2020-09-20 14:33:58Z

2

You need to build() the model with the appropriate input_shape parameter. This is what's causing this issue.

answered Sep 20, 2020 at 14:33

Christo S. Christov

2,2893 gold badges36 silver badges62 bronze badges

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

Comments

Mohammed Alali · Accepted Answer · 2022-08-30 23:17:43Z

2

I had this error and was able to solved it by adding a tf.keras.layers.InputLayer at the beginning of my model.

https://www.tensorflow.org/api_docs/python/tf/keras/layers/InputLayer

edited Aug 30, 2022 at 23:17

answered Dec 4, 2021 at 21:14

Mohammed Alali

713 bronze badges

Collectives™ on Stack Overflow

TF 2.3 Concatenate error - AttributeError: 'Sequential' object has no attribute '_nested_outputs'

2 Answers 2

Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related