Bug: TF Hub is not compatible with Keras 3 and TensorFlow 2.16+

[rkazants]

What happened?

When I try to create Keras model using TF 2.16+ I've got an error:
ValueError: Exception encountered when calling layer 'keras_layer' (type KerasLayer).
A KerasTensor is symbolic: it's a placeholder for a shape an a dtype. It doesn't have any actual numerical value. You cannot convert it to a NumPy array.

Relevant code

import tensorflow as tf
import tensorflow_hub as hub

image = tf.keras.layers.Input(shape=(224, 224, 3), dtype=tf.float32, name="image")
feature_vector = hub.KerasLayer("https://www.kaggle.com/models/google/mobilenet-v2/frameworks/tensorFlow2/variations/100-224-feature-vector/versions/2", trainable=False)(image)
softmax = tf.keras.layers.Dense(20, activation='softmax')(feature_vector)
classification_model = tf.keras.Model(inputs={'image': image}, outputs={'softmax': softmax})

Relevant log output

ValueError: Exception encountered when calling layer 'keras_layer' (type KerasLayer).
           A KerasTensor is symbolic: it's a placeholder for a shape an a dtype. It doesn't have any actual numerical value. You cannot convert it to a NumPy array.

tensorflow_hub Version

0.12.0 (latest stable release)

TensorFlow Version

other (please specify)

Other libraries

TF 2.16

Python Version

3.x

OS

Linux

[KeijiBranshi]

Hi @rkazants. Thanks for posting.

Can you try upgrading to the latest tensorflow_hub version 0.16.1 and installing tf-keras as a peer dependency?

Some extra context:

TensorFlow v2.16 points tf.keras to Keras 3, which unfortunately breaks a number of workflows with tensorflow_hub. We're working to make tensorflow_hub compatible with Keras 3 but in the meantime the recommendation is to use Keras 2 via tf-keras.