Tensorflow-examples: TypeError: unsupported operand type(s) for +: 'dict_values' and 'dict_values'

[Billy1900]

when I run this file tensorflow_v2/notebooks/3_NeuralNetworks/neural_network_raw.ipynb Link. the error occured, TypeError: unsupported operand type(s) for +: 'dict_values' and 'dict_values'.
How can I solve this?

my tf version 2.1.0, python 3.7

[lucianommartins]

@Billy1900 it happens because the notebooks were originally developed using Python 2.7.

So, if you run the notebook with a Python2.7 kernel, it will run flawlessly.

Also it fails on Python 3.7 due to some changes between Python2.7 and Python3.7:

• On Python2.7, dictionary.values() returns an object from list class;
• On Python3.7, dictionary.values() returns an object from dict_values class;

So to fix that problem, you can simply change line:

trainable_variables = weights.values() + biases.values()

by this content:

trainable_variables = list(weights.values()) + list(biases.values())

It solves the problem... Hope that helps.

[Billy1900]

@Billy1900 it happens because the notebooks were originally developed using Python 2.7.

So, if you run the notebook with a Python2.7 kernel, it will run flawlessly.

Also it fails on Python 3.7 due to some changes between Python2.7 and Python3.7:

• On Python2.7, dictionary.values() returns an object from list class;
• On Python3.7, dictionary.values() returns an object from dict_values class;

So to fix that problem, you can simply change line:

trainable_variables = weights.values() + biases.values()

by this content:

trainable_variables = list(weights.values()) + list(weights.values())

It solves the problem... Hope that helps.

trainable_variables = list(weights.values()) + list(weights.values()) or trainable_variables = list(weights.values()) + list(biass.values())?

[Billy1900]

it should be trainable_variables = list(weights.values()) + list(biass.values()) and it works, thanks!