max_trials variable in Tunner Class is set to None

[OmarIsmailAbdelrahman]

Describe the bug
When running the tuner component, it produces the following error: TypeError: '<' not supported between instances of 'int' and 'NoneType'. This issue is caused by the max_trials variable in the Oracle class inside the tuner. Upon tracing the code, I found that when Hyperband is initialized, it uses the HyperbandOracle class, which is a subclass of Engine.Oracle class. The problem appears to occur because max_trials is not set in the Oracle class. This results in an error when comparing len(sorted_trials) with max_trials in the Oracle file at line 677. Additionally, upon checking the super constructor of the Oracle class, it doesn't initialize a value for max_trials, which might be contributing to the issue.

To Reproduce
Provide steps to reproduce the bug: 1. Link: Colab Notebook that reproduces the issue.

Expected behavior
The tuner component should work without producing the error and handle the max_trials variable appropriately.

Additional context

i remember seeing it as a deprecated argument, and it happens on the latest release of keras

Would you like to help us fix it?
Yes

[edi-bice]

I have the same issue using the (keras) Tuner component in TFX 2.15

`import kerastuner

def _get_keras_hyperparameters() -> kerastuner.HyperParameters:
"""Returns hyperparameters for building Keras model."""
hp = kerastuner.HyperParameters()
# Defines search space.
hp.Choice('activation', ['relu', 'leaky_relu'])
hp.Choice('learning_rate', [1e-3, 1e-4, 1e-5], default=1e-4)
return hp

tuner = kerastuner.Hyperband(
hypermodel=_build_keras_model,
max_epochs=hparams.get('tune_epochs'), # Maximum number of epochs to train a model
factor=3, # Reduction factor for successive halving
hyperparameters=hparams,
tune_new_entries=True,
allow_new_entries=True,
objective=kerastuner.Objective('val_mean_absolute_error', 'min'),
directory=fn_args.working_dir,
project_name=configs.PIPELINE_NAME + '_tuning'
)`