How to execute TensorFlow 2 Keras Sequential model in eager mode when "compile"?

I want to build a loss in a "pythonic" way using the eager execution of TF2, but even in eager mode, Keras is passing non-eager tensors.

Code:

    def conditional_loss(self, y_true, y_pred):
        print(y_true)
        return 0

    def define_model(self):
        self.model = keras.Sequential([
            keras.layers.Dense(units=768),
            keras.layers.BatchNormalization(),
            keras.layers.ReLU(),
            keras.layers.Dropout(0.2),
            keras.layers.Dense(units=128),
            keras.layers.BatchNormalization(),
            keras.layers.ReLU(),
            keras.layers.Dropout(0.2),
            keras.layers.Dense(units=5, activation='softmax')
        ])

        self.model.compile(optimizer='adam',
                           loss=self.conditional_loss,
                           metrics=[self.conditional_loss, 
                                    keras.metrics.sparse_categorical_accuracy]
                           )
        self.model.fit(
            self.train_dataset,
            epochs=10,
            validation_data=self.test_dataset,
            callbacks=[tensorboard_callback, model_callback],
        )

If I print y_true in conditional_loss TF prints a non-eager tensor.

Tensor("metrics/conditional_loss/Cast:0", shape=(None, 1), dtype=float32)

If I build my own keras.Model() I can call it with the argumentdynamic=True to enable eager execution. (Reference). Exists a way to do it in keras.Sequential() ?

To do that you have to call model.compile() with the argument run_eagerly=True. Following the question example:

self.model.compile(optimizer='adam',
                           loss=self.conditional_loss,
                           metrics=[self.conditional_loss, 
                                    keras.metrics.sparse_categorical_accuracy],
                           run_eagerly=True
                           )