tensorflow: ValueError: Only instances of `keras.Layer` can be added to a Sequential mode

Issue type

Bug

Have you reproduced the bug with TensorFlow Nightly?

No

Source

source

TensorFlow version

v2.15.0-rc1-8-g6887368d6d4 2.15.0

Custom code

No

OS platform and distribution

Kaggle

Mobile device

No response

Python version

No response

Bazel version

No response

GCC/compiler version

No response

CUDA/cuDNN version

No response

GPU model and memory

No response

Current behavior?

When trying to build a sequential model with TFhub model USE in Kaggle, the notebook throws an error complaining that the hub model is not a valid class instance. At the same time, the same code works fine locally.

Standalone code to reproduce the issue

import pandas as pd
import tensorflow as tf
import tensorflow_hub as hub
from tensorflow.keras.optimizers import Adam, SGD
from tensorflow.keras.layers import Dense, Input, BatchNormalization, Dropout, Concatenate
from tensorflow.keras.models import Model, Sequential
from tensorflow.keras.callbacks import ModelCheckpoint
module_url = 'https://tfhub.dev/google/universal-sentence-encoder-large/5'
embed = hub.KerasLayer(module_url, trainable=True, name='USE_embedding')
def build_model(embed):
    
    model = Sequential([
        Input(shape=[], dtype=tf.string),
        embed,
        Dense(1, activation='sigmoid')
    ])
    model.compile(loss='mean_squared_error',
              optimizer=tf.keras.optimizers.Adam(
                  learning_rate=0.0001,
                  beta_1=0.9,
                  beta_2=0.999,
                  epsilon=1e-07,
                  amsgrad=False,
                  name='Adam'
              ), metrics=['accuracy'])
    return model


### Relevant log output

```shell
ValueError                                Traceback (most recent call last)
Cell In[9], line 1
----> 1 model=build_model(embed)
      2 model.fit(descriptions, labels, epochs=4)
      3 model.save('quality/use.keras')

Cell In[8], line 12, in build_model(embed)
     10 def build_model(embed):
---> 12     model = Sequential([
     13         Input(shape=[], dtype=tf.string),
     14         embed,
     15         Dense(1, activation='sigmoid')
     16     ])
     17     model.compile(loss='mean_squared_error',
     18               optimizer=tf.keras.optimizers.Adam(
     19                   learning_rate=0.0001,
   (...)
     24                   name='Adam'
     25               ), metrics=['accuracy'])
     26     return model

File /opt/conda/lib/python3.10/site-packages/keras/src/models/sequential.py:70, in Sequential.__init__(self, layers, trainable, name)
     68 if layers:
     69     for layer in layers:
---> 70         self.add(layer, rebuild=False)
     71     self._maybe_rebuild()

File /opt/conda/lib/python3.10/site-packages/keras/src/models/sequential.py:92, in Sequential.add(self, layer, rebuild)
     90         layer = origin_layer
     91 if not isinstance(layer, Layer):
---> 92     raise ValueError(
     93         "Only instances of `keras.Layer` can be "
     94         f"added to a Sequential model. Received: {layer} "
     95         f"(of type {type(layer)})"
     96     )
     97 if not self._is_layer_name_unique(layer):
     98     raise ValueError(
     99         "All layers added to a Sequential model "
    100         f"should have unique names. Name '{layer.name}' is already "
    101         "the name of a layer in this model. Update the `name` argument "
    102         "to pass a unique name."
    103     )

ValueError: Only instances of `keras.Layer` can be added to a Sequential model. Received: <tensorflow_hub.keras_layer.KerasLayer object at 0x793cedbef220> (of type <class 'tensorflow_hub.keras_layer.KerasLayer'>)

About this issue

  • Original URL
  • State: closed
  • Created 3 years ago
  • Comments: 26 (1 by maintainers)

Most upvoted comments

I got the same error when running the official notebook, https://www.tensorflow.org/tutorials/interpretability/integrated_gradients, the environment I am using is tensorflow==2.16 and tensorflow-hub==0.16. But if I roll back to tensorflow 2.14 and tensorflow-hub 0.15, there is no such error.

you have to wrap the hub layer in a Lambda layer like below:

import tensorflow as tf
import tensorflow_hub as hub

 # Wrap the hub layer in a Lambda layer

hub_layer_wrapper = tf.keras.layers.Lambda(lambda x: hub_layer(x))

model = tf.keras.Sequential([
    hub_layer_wrapper,  # Use the wrapped layer
    tf.keras.layers.Dense(16, activation='relu'),
    tf.keras.layers.Dense(1, activation='sigmoid')
])

model.summary()

I fix my code by using lambda layer I hope it is useful for you

fixed with:

!pip install tensorflow==2.15.0 tensorflow-hub keras==2.15.0

This is definitely a Tensorflow issue where the package management fails to check backward capability.

Consegui añadir el modelo con una lambda

import tensorflow_hub as hub

Cargar el modelo base preentrenado

base_model = hub.KerasLayer(“https://www.kaggle.com/models/google/mobilenet-v3/TensorFlow2/large-075-224-classification/1”, trainable=True, arguments=dict(batch_norm_momentum=0.997))

Construir el modelo

model = tf.keras.Sequential([ # Capa de entrada tf.keras.layers.InputLayer(input_shape=(224, 224, 3)),

# Envolver la capa del modelo base en una capa Lambda
tf.keras.layers.Lambda(lambda x: base_model(x)),

tf.keras.layers.Dropout(0.2),

# Capa densa final para clasificación
tf.keras.layers.Dense(3, activation='softmax')

])

Compilar el modelo

model.compile( optimizer=tf.keras.optimizers.Adam(learning_rate=0.0001), loss=‘sparse_categorical_crossentropy’, metrics=[‘accuracy’] )

Callbacks

early_stopping = tf.keras.callbacks.EarlyStopping(patience=10, restore_best_weights=True) reduce_lr = tf.keras.callbacks.ReduceLROnPlateau(factor=0.2, patience=5, min_lr=1e-6)

model.summary()

i had the same issue, tried all the things explained here but still having the same issue. i am not using kaggle i am coding on jupyter notebook. please help

i got the same issue also with Kaggle’s environment, with @SuryanarayanaY conclusion i got the solution by just pip install tensorflow again and found out that the keras library reinstalls to latest version

Hi @zzj0402 ,

The code executes fine on Colab as per gist. This moght be related to Kaggle environment which might have incompatible TF and tensorflow_hub version. You may check the Tensorflow and tensorflow_hub in kaggle environment and try to install those works in locally. Neverthless it’s not problem with Tensorflow or Keras.

Thanks!

you have to wrap the hub layer in a Lambda layer like below:

import tensorflow as tf
import tensorflow_hub as hub

 # Wrap the hub layer in a Lambda layer

hub_layer_wrapper = tf.keras.layers.Lambda(lambda x: hub_layer(x))

model = tf.keras.Sequential([
    hub_layer_wrapper,  # Use the wrapped layer
    tf.keras.layers.Dense(16, activation='relu'),
    tf.keras.layers.Dense(1, activation='sigmoid')
])

model.summary()

I fix my code by using lambda layer I hope it is useful for you

This worked. Although model.summary() not showing anything for layers. So, I believe this fix is just temporary and might get fixed on future versions.

fixed with:

!pip install tensorflow==2.15.0 tensorflow-hub keras==2.15.0

This is definitely a Tensorflow issue where the package management fails to check backward capability.

Thanks, this worked for me

Hi @SuryanarayanaY , you forgot to run build_model(embed) on your gist, when run, it will raise exactly the same error.

i cant even install older version of tensorflow the oldest i can install in 2.16