Added preprocessor and eval_batch_size key-word parameters to the inner_cv; added support of multi-inputs models; changed README and example/inner_cv

[dubovikmaster]

• Added the ability to use preprocessors of data such as keras.layers.Normalization or MinMaxScaler from sklearn.preprocessing for data preprocessing at each split into training and test samples during cross-validation. This allows the test to be closer to the actual use of the model.
• Support for cases where the model has multiple inputs.
• Added parameter eval_batch_size - allows you to control the size of the batch when validating the model on a test fold.
• Fixed typos in README.md.
• Added a description of new parameters to the README.md
• Added import of the inner_cv function and the OuterCV class to the __init__.py, which now allows you to do from keras_tuner_cv import inner_cv or OuterCV instead of from keras_tuner_cv.inner_cv import inner_cv
• The recently released version of keras-tuner 1.2 has incompatible changes. Therefore, I decided to limit the tuner version to version 1.1.3 with which keras-tuner-cv works.
• Added scikit-learn to the install requires

[dubovikmaster]

Hi @giuseppegrieco !
I hope you all are well!
In this pull request, I made a lot of fixes and improvements. Can you watch and accept it? Thank you very much for your work!

[Feheragyar]

else:
                    metrics = tuner_utils.convert_to_metrics_dict(
                        results, self.oracle.objective, "Tuner.run_trial()"
                    )
                    metrics.update(get_metrics_std_dict(results))
                    self.oracle.update_trial(
                        trial.trial_id,
                        metrics,
                    )

Small bug fix for inner_cv.py module:
The "Tuner.run_trial()" arg needs to be deleted for the metric conversion to work correctly, as it can only take 2 args.

[Furkan-rgb]

Hey guys, first of all thanks for the efforts. I'm running into a problem related to this and I hope someone can help me out as I'm still fairly new to keras_tuner. When trying to run inner_cv, After the first trial I get this error:

    metrics = tuner_utils.convert_to_metrics_dict(
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: convert_to_metrics_dict() takes 2 positional arguments but 3 were given

I have followed the examples but am unsure on what to do.
This is how I use it:

tuner = inner_cv(BayesianOptimization)(
    search_cnn_lstm_model,
    TimeSeriesSplit(n_splits=5),
    objective="val_loss",
    save_output=True,
    save_history=True,
    max_trials=max_trials,
    seed=42,
    executions_per_trial=2,
    directory="tmp/tb",
    project_name="cnn_lstm_vl_innercv",
)

tuner.search(
    X_train,
    y_train,
    epochs=120,
    shuffle=False,
    validation_data=(X_test, y_test),
    batch_size=72,
    callbacks=[
        tf.keras.callbacks.EarlyStopping(
            monitor="val_binary_accuracy", patience=10, mode="max"
        ),
    ],
    verbose=True,
)