CODEWITHSUNDEEP
tensorflowkerasconv-neural-networksequential
Marlon Hamm
Marlon Hamm

Reputation: 1

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

/usr/local/lib/python3.10/dist-packages/keras/layers/core/lambda_layer.py in_serialize_function_to_config(self, inputs, allow_raw)

    306             module = inputs.__module__
    307         elif callable(inputs):
--> 308             output = inputs.__name__
    309             output_type = "function"
    310             module = inputs.__module__

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

This is the error message I am looking at. I am doing transfer learning for a CNN (Xception) with some layers added before and after. Yesterday it trained without any issue, now the error above. I am training on google colab. I am using loss=keras.losses.BinaryCrossentropy(). For the 'lambda_layer' I used tf.keras.layers.Lambda(...).

Every idea is very welcome!

I used the 'Sequential' method at two places in my code, for both I already tried keras.models.Sequential(...) and tf.keras.Sequential(...), and ultmately pip installing tensorflow and keras again.

Upvotes: -1

Views: 283

Answers (1)

Marlon Hamm
Marlon Hamm

Reputation: 1

The error came from the following change of mine:

model_save = keras.callbacks.ModelCheckpoint('./Xception_model.h5',
                             save_best_only = True, 
                             save_weights_only = True,
                             monitor = 'val_loss',
                             mode = 'min', verbose = 1)

If in this callback save_weights_only is left out or changed to False, then the 'name' error occurs. This should not be the case at least according to the documentation.

Upvotes: 0

Related Questions