AttributeError: module 'tensorflow.keras.preprocessing' has no attribute 'image_dataset_from_directory' about keras-io HOT 16 CLOSED

Use the nightly build:

pip install tf-nightly

from keras-io.

Following the problem. I also installed tf-nightly, but still showing

AttributeError: module 'tensorflow_core.keras.preprocessing' has no attribute 'timeseries_dataset_from_array'

from keras-io.

Hi Team,
I am also having same issue, while running the example in tensorflow tutorials "Basic text classification" under "ML basics with Keras".

Traceback (most recent call last):
File "main.py", line 77, in
text_classification()
File "main.py", line 71, in text_classification
load_dataset(train_dir)
File "main.py", line 29, in load_dataset
raw_train_ds = tf.keras.preprocessing.text_dataset_from_directory(
AttributeError: module 'tensorflow.keras.preprocessing' has no attribute 'text_dataset_from_directory'

tensorflow version = 2.2.0
Python version = 3.6.9

I tried installing tf-nightly also. But it did not solve the issue.
Any more pointers to fix this issue.

from keras-io.

try tensorflow 2.3.0, that one contains text_dataset_from_directory ... at least on Windows

from keras-io.

I have solved this problem by pip install tf-nightly. If you did it but it ditn't take effect, you should restart your IDE or Jupyter notebook. I hope this comment is helpful for you!

from keras-io.

If you facing this problem now, upgrade to tensorflow 2.3.x. Also note that tf-nightly breaks your horovod.

from keras-io.

Same issue, tf-2.3.1

from keras-io.

FOR PEOPLE USING DOCKER TF FOR WSL2 and dont want to update their image:

1. Go to https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/keras/preprocessing
2. Download datasets_utils.py and text_dataset.py
3. make them as a module and call them within your project.

Note: text_datasets.py calls functions from datasets_utils.py. So you have to update the text_datasets.py file (Line 23) and make it import the datasets_utils.py properly.

from keras-io.

Use the nightly build:

pip install tf-nightly

After installing the nightly build in Kaggle, it still fails with no module found! Yet it works out of the box in Colab, why do we have all these inconsistencies? And how exactly can I use the image-dataset-from-directory function in Kaggle?

Has anyone successfully tried this in Kaggle?

Please I need to know urgently.

from keras-io.

Use the nightly build:

pip install tf-nightly

Thanks, but this didn't work for me.
I solved the problem with upgrading the TensorFlow, you may go with:
pip install --upgrade tensorflow

from keras-io.

pip install tf-nightly

RESTART
worked for me.

from keras-io.

I have created environment via "conda create workspace"

There I installed all necessary packages via conda, where tensorflow2.0.0 installed.

while importing

from Tensorflow.keras.layers import TextVectorization giving me import error that can not import TextVectorization and I'm getting the following error:
AttributeError: module 'tensorflow.keras.preprocessing' has no attribute 'image_dataset_from_directory'

when trying to solve: https://github.com/keras-team/keras-io/blob/master/examples/vision/image_classification_from_scratch.py

my TensorFlow version: 2.0.0
python version: 3.7.12

I tried pip3 install tf-nightly: got error zsh: illegal hardware instruction and module "tensorflow not found".

please help.

from keras-io.

Try with
keras.preprocessing.image_dataset.image_dataset_from_directory(

from keras-io.

Try using

tf.keras.applications.vgg16.preprocess_input(X, Dtype)
tf.keras.applications.resnet50.preprocess_input(x, data_format=None)

from keras-io.

make sure you are importing keras from tensorflow

import tensorflow as tf
from tensorflow import keras

from keras-io.

make sure you are importing keras from tensorflow

import tensorflow as tf from tensorflow import keras

This answer works for me.
my version is:
tf.version = '2.12.0'
keras.version = '2.12.0'

from keras-io.