Keras: Problem with visualization

[olaswietlicka]

Hi!

I have a problem with plot_model, which does not show my input layer on the plot.
Here is the code:

`from keras import models
from keras import layers
from keras.utils import plot_model

neural_network = models.Sequential()
neural_network.add(layers.Dense(units = 20,input_shape = (10,), activation = 'relu'))
neural_network.add(layers.Dense(units = 20, activation = 'sigmoid'))
neural_network.compile(loss = 'categorical_crossentropy', optimizer = 'sgd', metrics = ["accuracy"])
plot_model(neural_network, show_shapes=True, show_layer_names=True, to_file="neural_network.png")`

And it gives something like this:

[flyakite]

I found it's caused by omitting the first layer in Sequential in a637960fab61b66848a36e6a5caf0204c155af01

You can plot the model without using Sequential or remove the following lines in keras/engine/sequential.py

    @property
    def layers(self):
        # Historically, `sequential.layers` only returns layers that were added
        # via `add`, and omits the auto-generated `InputLayer`
        # that comes at the bottom of the stack.
        if self._layers and isinstance(self._layers[0], InputLayer):
            return self._layers[1:]
        return self._layers

[olaswietlicka]

Thank you.

[flyakite]

This may not be closed as this issue still exist. I only offered a temp solution. I'm not sure what's the historical reason we have to omits the InputLayer in the Sequential model.

[olundberg]

I still have this issue.

(base) C:WINDOWSsystem32>python -c "import keras; print(keras.__version__)"
Using TensorFlow backend.
2.2.2

Edit:
Tested with 2.2.3, still same error.

[b3326023]

I have this same issue.

[NicholaiStaalung]

Same issue

[endolith]

Same issue in tensorflow.keras. This also goes away if you use the functional API.

[ageron]

I submitted https://github.com/tensorflow/tensorflow/pull/24637 to fix this issue in tf.keras. Apparently it is already fixed in keras-team/keras. It's a trivial change (just one character added):

In the definition of model_to_dot() (in [tensorflow/python/]keras/utils/vis_utils.py), replace layers = model.layers with layers = model._layers.

[franzalex1980]

Hi, I am still very confused about getting this problem fixed! I am using keras = 2.2.4 on windows 10 Please, how do I define the input layer in the keras model to be visible in my plot_model function?
I would be so glad if someone helps.

Many thanks!

Alex

[dduemig]

In the definition of model_to_dot() (in [tensorflow/python/]keras/utils/vis_utils.py), replace layers = model.layers with layers = model._layers

does not work for me

[Walid-Ahmed]

I solved the issue by upgrading to Keras 2.3.1