keras 3.*, tensorflow 2.16 and keras 2.* compatibility problem

• TensorFlow Version: 2.16.1
• Keras Version: 3.3.3
• tf_keras Version: 2.16.0

Issue Description:

I encountered an error while running my training script training_holdout.py with the following command:

export TF_USE_LEGACY_KERAS=1 python3 training_holdout.py

Note: It's crucial to use export TF_USE_LEGACY_KERAS=1 to ensure that TensorFlow utilizes Keras version 2.* for compatibility reasons.

The script failed and produced the following traceback:

Traceback (most recent call last):  File "/media/fernando/B0EA304AEA300EDA/Dados/Fernando/CODE/PESQUISA/programs-tese/SYSTEM-TRAINING/cnn_emotion4/create/training_holdout.py", line 184, in <module>    model, target_size = mpp.create_model(model_type=model_type, load_weights=False)                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^  File "/media/fernando/B0EA304AEA300EDA/Dados/Fernando/CODE/PESQUISA/programs-tese/SYSTEM-TRAINING/cnn_emotion4/create/../library/BodyEmotion4Lib/lib_model.py", line 58, in create_model    modelo = tf.keras.Sequential([             ^^^^^^^^^^^^^^^^^^^^^  File "/home/fernando/.local/lib/python3.12/site-packages/tensorflow/python/trackable/base.py", line 204, in _method_wrapper    result = method(self, *args, **kwargs)             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^  File "/home/fernando/.local/lib/python3.12/site-packages/tf_keras/src/utils/traceback_utils.py", line 70, in error_handler    raise e.with_traceback(filtered_tb) from None  File "/usr/lib/python3.12/random.py", line 336, in randint    return self.randrange(a, b+1)           ^^^^^^^^^^^^^^^^^^^^^^  File "/usr/lib/python3.12/random.py", line 312, in randrange    istop = _index(stop)            ^^^^^^^^^^^^TypeError: 'float' object cannot be interpreted as an integer

Cause of the Error:

The error originates from the randint function in Python's random module, which is being passed a float instead of an integer. This causes a TypeError because the randrange function expects integer inputs.

Solution:

To resolve this issue, I modified line 336 in the file /usr/lib/python3.12/random.py from:

return self.randrange(a, b+1)

to:

return self.randrange(int(a), int(b+1))

This ensures that the arguments passed to randrange are integers, thereby preventing the TypeError.

Recommendation:

It would be beneficial to implement a type check or cast within the relevant parts of the TensorFlow or Keras codebase to handle such cases where float inputs might be passed inadvertently.

I would like to report this issue to the appropriate bug tracking system, but I am uncertain whether it falls under TensorFlow, Keras, or the Python random module. Could you please advise where to submit this bug report to ensure it reaches the correct library maintainers?