Question mark (?) in the model image using plot_model instead of None

[V2dha]

I am using the tf.keras.utils.plot_model function to plot the model. The code is given below. Now instead of getting None for the batch sizes I am getting question mark (?). So can you please address why this is happening and how I can replace None inplace of ? question mark.

tf.keras.utils.plot_model(
model,
to_file='model2.png',
show_shapes=True,
show_layer_names=True
)

[jpodivin]

What you see there is caused by this line in keras utils:

tensorflow/tensorflow/python/keras/utils/vis_utils.py

Line 183 in 2b96f36

return str(shape).replace(str(None), '?')

It's actually the desired behavior.
My understanding is that the None sized dimensions are effectively variable sized, or unknown '?' when the graph is compiled.
Therefore tensor with other dimensions matching can fit. Notice how output shape of a layer and the input shape of the following layer always matches.

Now you could replace the '?' string with 'None' string, but what would it be good for? It wouldn't make it any more readable.

[V2dha]

What you see there is caused by this line in keras utils:

tensorflow/tensorflow/python/keras/utils/vis_utils.py

Line 183 in 2b96f36

return str(shape).replace(str(None), '?')

It's actually the desired behavior.
My understanding is that the None sized dimensions are effectively variable sized, or unknown '?' when the graph is compiled.
Therefore tensor with other dimensions matching can fit. Notice how output shape of a layer and the input shape of the following layer always matches.

Now you could replace the '?' string with 'None' string, but what would it be good for? It wouldn't make it any more readable.

Actually I am working on a research paper in which I need the image of the model architecture although it wouldn't make it any more readable from a programmer's point of view but from a layman's understanding it would atleast come out to be a tad bit readable as the '?' mark signifies as if there's an error or something is missing where as None kind of explains that I voluntary didn't add the batch sizes. Thanks for the answer though!

[jpodivin]

Worst case scenario you can add an explanatory footnote.

[ravikyram]

@V2dha

It is expected behavior.Please, see this tutorial, link for your reference.Thanks!

[V2dha]

Okay thanks!