Keras: AttributeError: 'int' object has no attribute 'lower'

Question

I am building a neural network with keras and tensorfrolw as backend. it has 3 inputs from 0 to 9 and 3 outputs from 0 to 9. The data is served in a numpy array lie this: [ ['1' '4' '0'] ['6' '2' '1'] ...].

I am new to deep learning and this is one of my first neural networks so i am lost and have no idea what is causing this error.

I am aware that i probably need to change the optimizer, loss, metrics and probably more attributes, if anyone has any insights on that please share.

model = keras.Sequential([
    keras.layers.Flatten(3, input_shape=(3, 3)),
    keras.layers.Dense(9, activation="relu"),
    keras.layers.Dense(9, activation="relu"),
    keras.layers.Dense(3, activation="relu")
])
model.compile(optimizer="adam", loss="sparse_categorical_crossentropy", metrics=["accuracy"])
model.fit(training_input, training_output, epochs=5)

When i run this program i get this error:

Traceback (most recent call last):
  File "C:/Users/---/---/---/---/---/---/---/main.py", line 15, in <module>
    keras.layers.Flatten(3, input_shape=(3, 3)),
  File "C:\Users\---\AppData\Local\Programs\Python\Python37\lib\site-packages\tensorflow_core\python\keras\layers\core.py", line 571, in __init__
    self.data_format = conv_utils.normalize_data_format(data_format)
  File "C:\Users\---\AppData\Local\Programs\Python\Python37\lib\site-packages\tensorflow_core\python\keras\utils\conv_utils.py", line 191, in normalize_data_format
    data_format = value.lower()
AttributeError: 'int' object has no attribute 'lower' \

duhaime · Accepted Answer · 2020-05-28 01:19:58Z

1

I had provided an argument to the flatten layer when I shouldn't have. Busted:

import tensorflow.keras.layers as l

h = i = l.Input(shape=(10, 5))
h = l.Flatten(50)(h) # don't do this!
o = l.Dense(50)(h)
model = keras.Model(inputs=i, outputs=o)
model.compile(optimizer='adam', loss='mse')

With no argument specified the model compiles:

import tensorflow.keras.layers as l

h = i = l.Input(shape=(10, 5))
h = l.Flatten()(h) # do this!
o = l.Dense(50)(h)
model = keras.Model(inputs=i, outputs=o)
model.compile(optimizer='adam', loss='mse')

answered May 28, 2020 at 1:19

duhaime

27.9k22 gold badges196 silver badges243 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Physicing · Accepted Answer · 2019-10-20 10:54:21Z

0

Please check Keras Documentation for using Flatten() layer. You are misusing the Flatten. I suggest you to do the reshaping of your data before your model and feed forward into the model.

answered Oct 20, 2019 at 10:54

Physicing

5425 silver badges17 bronze badges

2 Comments

A45 Over a year ago

Thanks dude. I should have written keras.layers.Dense(3, input_shape=(3,)),

Physicing Over a year ago

Hope your problem resolves. Seems like your optimizers, loss or metrics is not problematic.

Collectives™ on Stack Overflow

Keras: AttributeError: 'int' object has no attribute 'lower'

2 Answers 2

Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related