def write_log(callback, names, logs, batch_no):
            for name, value in zip(names, logs):
                summary = tf.Summary()
                summary_value = summary.value.add()
                summary_value.simple_value = value
                summary_value.tag = name
                callback.writer.add_summary(summary, batch_no)
                callback.writer.flush()

        callback = callbacks[0]
        callback.set_model(self.keras_model)
        train_names = ['train_loss', 'train_mae']
        val_names = ['val_loss', 'val_mae']

        for epoch in range(epochs):
            print("epoch: {}".format(epoch))
            for step in range(self.config.STEPS_PER_EPOCH):
                # print("step: {}".format(step))
                if step % int(self.config.STEPS_PER_EPOCH / 100.) == 0:
                    print('.', end='', flush=True)
                train_inputs = next(train_generator)
                logs = self.keras_model.train_on_batch(train_inputs[0], train_inputs[1])
                write_log(callback, train_names, logs, step + epoch * self.config.STEPS_PER_EPOCH)

                if step % 50 == 0:
                    val_inputs = next(val_generator)
                    logs = self.keras_model.train_on_batch(val_inputs[0], val_inputs[1])
                    write_log(callback, val_names, logs, step + epoch * self.config.STEPS_PER_EPOCH)
            print('')  # newline
            self.keras_model.save(os.path.join(self.log_dir, 'tiny_{}.h5'.format(epoch)))

        self.epoch = max(self.epoch, epochs)