Error with tensorflow 3.13

[sylvchev]

When upgrading to TF 3.13, the documentation building raise the following error :

In [2]: import tensorflow as tf
   ...: print(tf.reduce_sum(tf.random.normal([1000, 1000])))
2023-10-12 18:24:11.248019: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
tf.Tensor(1321.248, shape=(), dtype=float32)

In [3]: import moabb.pipelines.deep_learning
Tensorflow not install, you could not use those pipelines
---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
Cell In[3], line 1
----> 1 import moabb.pipelines.deep_learning

File ~/src/github/moabb-sylvchev/moabb/pipelines/deep_learning.py:32
     14 from keras.constraints import max_norm
     15 from keras.layers import (
     16     Activation,
     17     Add,
   (...)
     30     Permute,
     31 )
---> 32 from keras.layers.normalization.batch_normalization import BatchNormalization
     33 from keras.models import Model, Sequential
     34 from scikeras.wrappers import KerasClassifier

ModuleNotFoundError: No module named 'keras.layers.normalization'

You could see that MOABB indicates that Tensorflow not install, you could not use those pipelines even when tensorflow is correctly installed and working in the same environment.

Downgrading to tensorflow 3.12 solve the issue. @carraraig Do you have any idea why?

[brianjohannes]

I found a potential solution to this here: https://stackoverflow.com/questions/69471749/importerror-cannot-import-name-batchnormalization-from-keras-layers-normaliz

The solution is to import BatchNormalization from keras.layers

from tensorflow.keras.layers import BatchNormalization