ValueError: Only instances of keras.Layer can be added to a Sequential model. Received: <tensorflow_hub.keras_layer.KerasLayer object at 0x7a9a39daf130> (of type <class 'tensorflow_hub.keras_layer.KerasLayer'>)

[moatednorth]

Discussed in #662

^{Originally posted by moatednorth July 31, 2024}
Sequential api can't handle hub.KerasLayer

`

Create a Keras Layer using the USE pretrained layer from Kaggle

sentence_encoder_layer = hub.KerasLayer("https://www.kaggle.com/models/google/universal-sentence-encoder/TensorFlow2/universal-sentence-encoder/2",
input_shape=[],
dtype=tf.string,
trainable=False,
name="USE")

Create model using the Sequential API

model_6 = tf.keras.Sequential([
sentence_encoder_layer,
layers.Dense(64, activation="relu"),
layers.Dense(1, activation="sigmoid", name="ouput_layer")
], name="model_6_USE")
`

[AnilKamath27]

Hi, moatednorth. This is an issue that is occurring in the latest version of tensorflow 2.17. This won't be an issue if you try with tf 2.15.0. Below is a screenshot of my notebook where I was able to use hub.KerasLayers as part of Sequential API.

[ctournas]

I just stuck in this for an hour, for anyone out there that stuck on this too... just run this: pip install tensorflow==2.15.0 tensorflow-hub keras==2.15.0

[GoJo-Rika]

I am getting same issue with TensorFlow == 2.18.0

  IMAGE_SHAPE = (224, 224)
  # Download the pretrained model and save it as a Keras layer
  resnet_url = "https://tfhub.dev/google/imagenet/resnet_v2_50/feature_vector/4"
  feature_extractor_layer = hub.KerasLayer(resnet_url,
                                            trainable=False, # freeze the already learned patterns
                                            name="feaure_extraction_layer",
                                            input_shape=IMAGE_SHAPE+(3,))
  
  # Create our own model
  resnet_model = tf.keras.Sequential([
      feature_extractor_layer,
      layers.Dense(10, activation="softmax", name="output_layer")
  ])

Getting a ValueError, here is a detailed explanaion:

ValueError: Only instances of `keras.Layer` can be added to a Sequential model. Received: <tensorflow_hub.keras_layer.KerasLayer object at 0x7d6a32dac1d0> (of type <class 'tensorflow_hub.keras_layer.KerasLayer'>)

I was able to resolve the above issue by wrapping the hub layer in the Lambda layer like below:

IMAGE_SHAPE = (224, 224)
# Download the pretrained model and save it as a Keras layer
resnet_url = "https://tfhub.dev/google/imagenet/resnet_v2_50/feature_vector/4"
feature_extractor_layer = hub.KerasLayer(resnet_url,
                                          trainable=False, # freeze the already learned patterns
                                          name="feaure_extraction_layer",
                                          input_shape=IMAGE_SHAPE+(3,))

# Create our own model
resnet_model = tf.keras.Sequential([
    tf.keras.layers.Lambda(lambda x: feature_extractor_layer(x)),
    layers.Dense(10, activation="softmax", name="output_layer")
])

But model.summary() doesn't give any details from feature_extractor_layer

Some URLS that I found useful:

• ValueError: Only instances of keras.Layer can be added to a Sequential mode tensorflow/tensorflow#63085 (comment)
• https://stackoverflow.com/questions/78530756/error-only-instances-of-keras-layer-can-be-added-to-a-sequential-model - this url gives a sligtly different solution.

Read a couple of GitHub issues discussions and also a Medium article on the same and they resolved the above problem using the 2nd option. But I didn't tried it myself as the first option worked for me.

Here are the version of tensorflow, tensorflow_hub and keras that I was using:

tensorflow == 2.18.0
tensorflow_hub == 0.16.1
keras == 3.8.0
tf_keras == 2.18.0 (2nd solution)