Too restrictive ModuleNotFoundError exception in convolutional

[Algue-Rythme]

Version

Tensorflow: 2.13
Deel.lip 1.4.0

Issue

In file : https://github.com/deel-ai/deel-lip/blob/master/deel/lip/layers/convolutional.py

There is the code:

try:
    from keras.utils import conv_utils  # in Keras for TF >= 2.6
except ModuleNotFoundError:
    from tensorflow.python.keras.utils import conv_utils  # in TF.python for TF <= 2.5

In tensorflow>=2.13 the new path for conv_utils is tensorflow.python.keras.utils. Unfortunately, attempting to import with from keras.utils import conv_utils does not yield a ModuleNotFoundError becasue the module exists: instead, it simply yield a ImportError.

Since ImportError is not ModuleNotFoundError the except does not catch anything.

Solution

However observe that per this post that ModuleNotFoundError is an instance of ImportError , therefore we should replace the exception with a more general one.

[cofri]

Hi @Algue-Rythme,
TensorFlow indeed changed (again) the way conv_utils is imported... @danibene suggested in a previous PR to fix this issue (04b03f8). His commit was integrated in a broader PR #76 with corrections for other failures introduced by TF 2.13.

However, we didn't notice that ModuleNotFoundError is a subclass of ImportError. We can simplify the open PR #76 with your proposition.

[thib-s]

Closing this issue as the suggested fix has been merged in #76 . Thank for the contribution !