search

keras-team / autokeras

AutoML library for deep learning
http://autokeras.com/
Apache License 2.0
9.1k stars 1.4k forks source link

Bug: ModuleNotFoundError: No module named 'tensorflow.keras.layers.experimental' #1923

Open jordannelson0 opened 1 week ago

jordannelson0 commented 1 week ago

Bug Description

I am getting this error: File "C:\Users\jorda\Documents\PhD\Project\NLP_algorithms\CNN_AM.py", line 3, in import autokeras as ak File "C:\Users\jorda\Documents\PhD\Project\venv\Lib\site-packages\autokeras\__init__.py", line 15, in from autokeras.auto_model import AutoModel File "C:\Users\jorda\Documents\PhD\Project\venv\Lib\site-packages\autokeras\auto_model.py", line 26, in from autokeras import blocks File "C:\Users\jorda\Documents\PhD\Project\venv\Lib\site-packages\autokeras\blocks\__init__.py", line 18, in from autokeras.blocks.basic import BertBlock File "C:\Users\jorda\Documents\PhD\Project\venv\Lib\site-packages\autokeras\blocks\basic.py", line 25, in from autokeras import keras_layers File "C:\Users\jorda\Documents\PhD\Project\venv\Lib\site-packages\autokeras\keras_layers.py", line 27, in from tensorflow.keras.layers.experimental import preprocessing ModuleNotFoundError: No module named 'tensorflow.keras.layers.experimental' ### Bug Reproduction Code for reproducing the bug: `import os import autokeras as ak import numpy as np from pandas import read_csv from sklearn.model_selection import train_test_split os.environ["TF_ENABLE_ONEDNN_OPTS"] = "0" dataframe = read_csv("ML_Prompts.csv", skiprows=0) dataset = dataframe.values x = dataset[:, 0:1744] y = dataset[:, 1744] x_train, x_temp, y_train, y_temp = train_test_split(x, y, test_size=0.2) x_val, x_test, y_val, y_test = train_test_split(x_temp, y_temp, test_size=0.5) x_train = np.expand_dims(x_train, axis=2) input_node = ak.Input() output_node = ak.ConvBlock()(input_node) output_node = ak.SpatialReduction()(input_node) # optional output_node = ak.ConvBlock()(input_node) output_node = ak.SpatialReduction()(output_node) # optional output_node = ak.ClassificationHead(num_classes=8, multi_label=False)(output_node) model = ak.AutoModel( inputs=input_node, outputs=output_node, overwrite=True, max_trials=20, objective='val_accuracy', tuner='bayesian' ) model.fit(x_train, y_train, epochs=50, batch_size=4, verbose=1) eval_result = model.evaluate(x_test, y_test, verbose=1) print("Test loss:", eval_result[0]) print("Test accuracy:", eval_result[1]) modelv = model.export_model() print(type(model.summary())) try: modelv.save("CNN_ML_Classifier.keras") except Exception: modelv.save("CNN_ML_Classifier.keraas") ` Data used by the code: ### Expected Behavior

Setup Details

Include the details about the versions of:

  • OS type and version: Windows 11
  • Python: 3.12.4
  • autokeras: 1.0.20 (I have been unable to upgrade this package due to a different error)
  • keras-tuner: 1.4.7
  • scikit-learn: 1.5.0
  • numpy: 1.26.4
  • pandas: 2.2.2
  • tensorflow: 2.17.0rc0

Additional context

Daniel-Alvarenga commented 1 week ago

The error you're encountering (ModuleNotFoundError: No module named 'tensorflow.keras.layers.experimental') suggests that the version of TensorFlow you are using doesn't have the tensorflow.keras.layers.experimental module. This module has been deprecated in the latest versions of TensorFlow.

To resolve this, you can try this:

Check TensorFlow and AutoKeras Compatibility

First, ensure that your versions of TensorFlow and AutoKeras are compatible. AutoKeras 1.0.20 might not be fully compatible with TensorFlow 2.17.0rc0. You can check the compatibility matrix on the AutoKeras documentation or GitHub repository.

Update or Downgrade TensorFlow

You may need to downgrade TensorFlow to a version compatible with AutoKeras 1.0.20. For example, TensorFlow 2.4.1 is known to be compatible with AutoKeras 1.0.20. You can downgrade TensorFlow using pip:

pip install tensorflow==2.4.1

Install the Latest AutoKeras

If downgrading TensorFlow is not an option, you may need to upgrade AutoKeras. However, since you mentioned an error while upgrading AutoKeras, we'll address that next.

Resolve AutoKeras Upgrade Issue

Try upgrading AutoKeras to the latest version using pip:

pip install --upgrade autokeras

If you encounter errors during the upgrade, you can try uninstalling and then reinstalling it:

pip uninstall autokeras
pip install autokeras
jordannelson0 commented 1 week ago

The error you're encountering (ModuleNotFoundError: No module named 'tensorflow.keras.layers.experimental') suggests that the version of TensorFlow you are using doesn't have the tensorflow.keras.layers.experimental module. This module has been deprecated in the latest versions of TensorFlow.

To resolve this, you can try this:

Check TensorFlow and AutoKeras Compatibility

First, ensure that your versions of TensorFlow and AutoKeras are compatible. AutoKeras 1.0.20 might not be fully compatible with TensorFlow 2.17.0rc0. You can check the compatibility matrix on the AutoKeras documentation or GitHub repository.

Update or Downgrade TensorFlow

You may need to downgrade TensorFlow to a version compatible with AutoKeras 1.0.20. For example, TensorFlow 2.4.1 is known to be compatible with AutoKeras 1.0.20. You can downgrade TensorFlow using pip:

pip install tensorflow==2.4.1

Install the Latest AutoKeras

If downgrading TensorFlow is not an option, you may need to upgrade AutoKeras. However, since you mentioned an error while upgrading AutoKeras, we'll address that next.

Resolve AutoKeras Upgrade Issue

Try upgrading AutoKeras to the latest version using pip:

pip install --upgrade autokeras

If you encounter errors during the upgrade, you can try uninstalling and then reinstalling it:

pip uninstall autokeras
pip install autokeras

I can't get autokeras 2.0 because of this: The conflict is caused by: keras-nlp 0.14.0 depends on tensorflow-text; platform_system != "Darwin" keras-nlp 0.12.1 depends on tensorflow-text; platform_system != "Darwin" keras-nlp 0.12.0 depends on tensorflow-text; platform_system != "Darwin" keras-nlp 0.11.1 depends on tensorflow-text; platform_system != "Darwin" keras-nlp 0.11.0 depends on tensorflow-text; platform_system != "Darwin" keras-nlp 0.10.0 depends on tensorflow-text; platform_system != "Darwin" keras-nlp 0.9.3 depends on tensorflow-text; platform_system != "Darwin" keras-nlp 0.9.2 depends on tensorflow-text; platform_system != "Darwin" keras-nlp 0.9.1 depends on tensorflow-text; platform_system != "Darwin" keras-nlp 0.9.0 depends on tensorflow-text; platform_system != "Darwin" keras-nlp 0.8.2 depends on tensorflow-text; platform_system != "Darwin" keras-nlp 0.8.1 depends on tensorflow-text; platform_system != "Darwin" keras-nlp 0.8.0 depends on tensorflow-text; platform_system != "Darwin"