How to use the alibi.utils.tf._check_keras_or_tf function in alibi

clip
            Tuple with min and max clip ranges for both the numerical gradients and the gradients
            obtained from the TensorFlow graph
        update_num_grad
            If numerical gradients are used, they will be updated every update_num_grad iterations
        no_info_val
            Global or feature-wise value considered as containing no information
        write_dir
            Directory to write tensorboard files to
        sess
            Optional Tensorflow session that will be used if passed instead of creating or inferring one internally
        """
        self.predict = predict

        # check whether the model and the auto-encoder are Keras or TF models and get session
        is_model, is_model_keras, model_sess = _check_keras_or_tf(predict)
        is_ae, is_ae_keras, ae_sess = _check_keras_or_tf(ae_model)
        # TODO: check ae and model are compatible

        # if session provided, use it
        if isinstance(sess, tf.compat.v1.Session):
            self.sess = sess
        else:
            self.sess = model_sess

        if is_model:  # Keras or TF model
            self.model = True
            classes = self.sess.run(self.predict(tf.convert_to_tensor(np.zeros(shape), dtype=tf.float32))).shape[1]
        else:
            self.model = False
            classes = self.predict(np.zeros(shape)).shape[1]

eps[1] it should be (1x nb of features)
        clip
            Tuple with min and max clip ranges for both the numerical gradients and the gradients
            obtained from the TensorFlow graph
        update_num_grad
            If numerical gradients are used, they will be updated every update_num_grad iterations
        write_dir
            Directory to write tensorboard files to
        sess
            Optional Tensorflow session that will be used if passed instead of creating or inferring one internally
        """
        self.predict = predict

        # check whether the model, encoder and auto-encoder are Keras or TF models and get session
        is_model, is_model_keras, model_sess = _check_keras_or_tf(predict)
        is_ae, is_ae_keras, ae_sess = _check_keras_or_tf(ae_model)
        is_enc, is_enc_keras, enc_sess = _check_keras_or_tf(enc_model)
        # TODO: check ae, enc and model are all compatible

        # if session provided, use it
        if isinstance(sess, tf.compat.v1.Session):
            self.sess = sess
        else:
            self.sess = model_sess

        if is_model:  # Keras or TF model
            self.model = True
            self.classes = self.predict.predict(np.zeros(shape)).shape[1]  # type: ignore
        else:  # black-box model
            self.model = False
            self.classes = self.predict(np.zeros(shape)).shape[1]

Tuple with min and max clip ranges for both the numerical gradients and the gradients
            obtained from the TensorFlow graph
        update_num_grad
            If numerical gradients are used, they will be updated every update_num_grad iterations
        no_info_val
            Global or feature-wise value considered as containing no information
        write_dir
            Directory to write tensorboard files to
        sess
            Optional Tensorflow session that will be used if passed instead of creating or inferring one internally
        """
        self.predict = predict

        # check whether the model and the auto-encoder are Keras or TF models and get session
        is_model, is_model_keras, model_sess = _check_keras_or_tf(predict)
        is_ae, is_ae_keras, ae_sess = _check_keras_or_tf(ae_model)
        # TODO: check ae and model are compatible

        # if session provided, use it
        if isinstance(sess, tf.compat.v1.Session):
            self.sess = sess
        else:
            self.sess = model_sess

        if is_model:  # Keras or TF model
            self.model = True
            classes = self.sess.run(self.predict(tf.convert_to_tensor(np.zeros(shape), dtype=tf.float32))).shape[1]
        else:
            self.model = False
            classes = self.predict(np.zeros(shape)).shape[1]

        self.mode = mode

clip
            Tuple with min and max clip ranges for both the numerical gradients and the gradients
            obtained from the TensorFlow graph
        update_num_grad
            If numerical gradients are used, they will be updated every update_num_grad iterations
        write_dir
            Directory to write tensorboard files to
        sess
            Optional Tensorflow session that will be used if passed instead of creating or inferring one internally
        """
        self.predict = predict

        # check whether the model, encoder and auto-encoder are Keras or TF models and get session
        is_model, is_model_keras, model_sess = _check_keras_or_tf(predict)
        is_ae, is_ae_keras, ae_sess = _check_keras_or_tf(ae_model)
        is_enc, is_enc_keras, enc_sess = _check_keras_or_tf(enc_model)
        # TODO: check ae, enc and model are all compatible

        # if session provided, use it
        if isinstance(sess, tf.compat.v1.Session):
            self.sess = sess
        else:
            self.sess = model_sess

        if is_model:  # Keras or TF model
            self.model = True
            self.classes = self.predict.predict(np.zeros(shape)).shape[1]  # type: ignore
        else:  # black-box model
            self.model = False
            self.classes = self.predict(np.zeros(shape)).shape[1]

        if is_enc:

numpy arrays. For eps[0], the array dimension should be (1x nb of prediction categories) and for
            eps[1] it should be (1x nb of features)
        clip
            Tuple with min and max clip ranges for both the numerical gradients and the gradients
            obtained from the TensorFlow graph
        update_num_grad
            If numerical gradients are used, they will be updated every update_num_grad iterations
        write_dir
            Directory to write tensorboard files to
        sess
            Optional Tensorflow session that will be used if passed instead of creating or inferring one internally
        """
        self.predict = predict

        # check whether the model, encoder and auto-encoder are Keras or TF models and get session
        is_model, is_model_keras, model_sess = _check_keras_or_tf(predict)
        is_ae, is_ae_keras, ae_sess = _check_keras_or_tf(ae_model)
        is_enc, is_enc_keras, enc_sess = _check_keras_or_tf(enc_model)
        # TODO: check ae, enc and model are all compatible

        # if session provided, use it
        if isinstance(sess, tf.compat.v1.Session):
            self.sess = sess
        else:
            self.sess = model_sess

        if is_model:  # Keras or TF model
            self.model = True
            self.classes = self.predict.predict(np.zeros(shape)).shape[1]  # type: ignore
        else:  # black-box model
            self.model = False
            self.classes = self.predict(np.zeros(shape)).shape[1]

# options for the optimizer
        self.max_iter = max_iter
        self.lam_init = lam_init
        self.tol = tol
        self.max_lam_steps = max_lam_steps
        self.early_stop = early_stop

        self.eps = eps
        self.init = init
        self.feature_range = feature_range
        self.target_proba_arr = target_proba * np.ones(self.batch_size)

        self.debug = debug

        # check if the passed object is a model and get session
        is_model, is_keras, model_sess = _check_keras_or_tf(predict_fn)

        # if session provided, use it
        if isinstance(sess, tf.compat.v1.Session):
            self.sess = sess
        else:
            self.sess = model_sess

        if is_model:  # Keras or TF model
            self.model = True
            self.predict_fn = predict_fn.predict  # type: ignore # array function
            self.predict_tn = predict_fn  # tensor function

        else:  # black-box model
            self.predict_fn = predict_fn
            self.predict_tn = None
            self.model = False