Test failing with tensorflow 2.17 #96
Description
System information
- Windows 11
- Python 3.11.9
- DeepTables 0.2.6
- Tensorflow 2.17.0
Describe the current behavior
Running the standard test snippet, as well as trying to train a model, fails on:
File "C:\Me\Epoch\Code\mcts\.venv\Lib\site-packages\deeptables\models\deeptable.py", line 363, in fit
history = model.fit(X, y, batch_size=batch_size, epochs=epochs, verbose=verbose, shuffle=shuffle,
history = model.fit(X, y, batch_size=batch_size, epochs=epochs, verbose=verbose, shuffle=shuffle,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Me\Epoch\Code\mcts\.venv\Lib\site-packages\deeptables\models\deepmodel.py", line 106, in fit
history = self.model.fit(train_data,
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Me\Epoch\Code\mcts\.venv\Lib\site-packages\keras\src\utils\traceback_utils.py", line 122, in error_handler
raise e.with_traceback(filtered_tb) from None
File "C:\Me\Epoch\Code\mcts\.venv\Lib\site-packages\keras\src\utils\traceback_utils.py", line 117, in error_handler
return fn(*args, **kwargs)
^^^^^^^^^^^^^
TypeError: TensorFlowTrainer.fit() got an unexpected keyword argument 'max_queue_size'
Standalone code to reproduce the issue
python -c "from deeptables.utils.quicktest import test; test()"
Suspected cause
max_queue_size ,workers, and use_multiprocessing are passed. These appear to not exist in tensorflow 2.17, although they did in older versions. Downgrading to tensorflow==2.15.0 made the test pass. I am reporting this as a bug as there is no constraint mentioned anywhere about supported versions, only >2.4 in the requirements.txt.