tensorflow keras: I am getting this error 'module "tensorflow._api.v1.keras.layers' has no attribute 'flatten'"

Question

I am getting the above error while executing the below code. I am trying to work out this below tutorial on tensorflow neural network implementation. https://www.datacamp.com/community/tutorials/tensorflow-tutorial

def load_data(data_directory):
directories = [d for d in os.listdir(data_directory) 
               if os.path.isdir(os.path.join(data_directory, d))]
labels = []
images = []
for d in directories:
    label_directory = os.path.join(data_directory, d)
    file_names = [os.path.join(label_directory, f) 
                  for f in os.listdir(label_directory) 
                  if f.endswith(".ppm")]
    for f in file_names:
        images.append(skimage.data.imread(f))
        labels.append(int(d))
return images, labels

import os
import skimage
from skimage import transform
from skimage.color import rgb2gray
import numpy as np
import keras
from keras import layers
from keras.layers import Dense
ROOT_PATH = "C://Users//Jay//AppData//Local//Programs//Python//Python37//Scriptcodes//BelgianSignals"
train_data_directory = os.path.join(ROOT_PATH, "Training")
test_data_directory = os.path.join(ROOT_PATH, "Testing")

images, labels = load_data(train_data_directory)


# Print the `labels` dimensions
print(np.array(labels))

# Print the number of `labels`'s elements
print(np.array(labels).size)

# Count the number of labels
print(len(set(np.array(labels))))

# Print the `images` dimensions
print(np.array(images))

# Print the number of `images`'s elements
print(np.array(images).size)

# Print the first instance of `images`
np.array(images)[0]

images28 = [transform.resize(image, (28, 28)) for image in images]

images28 = np.array(images28)

images28 = rgb2gray(images28)

# Import `tensorflow` 
import tensorflow as tf 

# Initialize placeholders 
x = tf.placeholder(dtype = tf.float32, shape = [None, 28, 28])
y = tf.placeholder(dtype = tf.int32, shape = [None])

# Flatten the input data
images_flat = tf.keras.layers.flatten(x)

# Fully connected layer 
logits = tf.contrib.layers.dense(images_flat, 62, tf.nn.relu)

# Define a loss function
loss = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(labels = y, 
                                                                    logits = logits))
# Define an optimizer 
train_op = tf.train.AdamOptimizer(learning_rate=0.001).minimize(loss)

# Convert logits to label indexes
correct_pred = tf.argmax(logits, 1)

# Define an accuracy metric
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))

At first, I used tf.layers.flatten(x) as in the tutorial. however, it will be depreciated in future versions. So add keras instead as suggested.

I am getting the following output in IDLE Console.

RESTART: C:\Users\Jay\AppData\Local\Programs\Python\Python37\Scriptcodes\SecondTensorFlow.py Using TensorFlow backend.

Warning (from warnings module): File "C:\Users\Jay\AppData\Local\Programs\Python\Python37\lib\site-packages\skimage\transform_warps.py", line 105 warn("The default mode, 'constant', will be changed to 'reflect' in " UserWarning: The default mode, 'constant', will be changed to 'reflect' in skimage 0.15.

Warning (from warnings module): File "C:\Users\Jay\AppData\Local\Programs\Python\Python37\lib\site-packages\skimage\transform_warps.py", line 110 warn("Anti-aliasing will be enabled by default in skimage 0.15 to " UserWarning: Anti-aliasing will be enabled by default in skimage 0.15 to avoid aliasing artifacts when down-sampling images.

Traceback (most recent call last): File "C:\Users\Jay\AppData\Local\Programs\Python\Python37\Scriptcodes\SecondTensorFlow.py", line 64, in

images_flat = tf.python.keras.layers.flatten(x)

AttributeError: module 'tensorflow' has no attribute 'python'

I am using, Keras version 2.2.4 Tensorflow version 1.13.1

Answer 1

Either

from keras.layers import Flatten

and use

Flatten()(input)

or

simply use

tf.keras.layers.Flatten()(input)

Answer 2

The new ("keras as the default API") approach would have you use the keras layer tf.keras.layers.Flatten but there is a little nuance you seem to have missed (and that hasn't been mentioned in the comments).

tf.keras.layers.Flatten() actually returns a keras layer (callable) object which in turn needs to be called with your previous layer.

So something more like this:

# Flatten the input data
flatten_layer = tf.keras.layers.Flatten()
images_flat = flatten_layer(x)

or, for brevity, just:

# Flatten the input data
images_flat = tf.keras.layers.Flatten()(x)