图片生成器未设置功能性Keras模型中的Validation_data

Question

我正在尝试在Keras中构建自定义回调，以跟踪精度并在每个时期结束时进行调用。

class Precision(Callback):
    """
    Keras Callback. Calculates precision metrics at the end of each epoch.
    """
    def __init__(self):
        super().__init__()
        self.precisions = []

    def on_train_begin(self, logs={}):
        self.precisions = []

    def on_epoch_end(self, epoch, logs={}):
        y_pred = (np.asarray(self.model.predict(self.validation_data[0]))).round()
        y_true = self.validation_data[1]
        precision = precision_score(y_true, y_pred)
        self.precisions.append(precision)
        print("validation set precision at epoch {}: {}".format(epoch, precision))
        return
# ------------
model.fit_generator(training_generator,
                    steps_per_epoch=(TRAIN_SIZE / BATCH_SIZE),  # number of samples in the dataset
                    epochs=EPOCHS,  # number of epochs, training cycles
                    validation_data=validation_generator,  # performance eval on test set
                    validation_steps=(TEST_SIZE / BATCH_SIZE),
                    callbacks=[history,
                               precision])

我收到以下错误：

File "/Users/user/Desktop/cnn_toolkit.py", line 68, in on_epoch_end
    y_pred = (np.asarray(self.model.predict(self.validation_data[0]))).round()
TypeError: 'NoneType' object is not subscriptable

我正在做迁移学习（Inception V3），并使用功能性Keras Model对象。我尝试更新tensorflow，从源代码构建它，还尝试了self.model.validation_data仍然不起作用。到处都在寻找答案，却找不到能解决我问题的任何东西。我有Keras 2.2.4。