嘗試訓練一個 Robust CNN 模型，其定義如下：from keras.datasets import cifar10from keras.utils import np_utilsfrom keras import metricsfrom keras.models import Sequentialfrom keras.layers import Dense, Flatten, Conv2D, MaxPooling2D, LSTM, mergefrom keras.layers import BatchNormalizationfrom keras import metricsfrom keras.losses import categorical_crossentropyfrom keras.optimizers import SGDimport pickleimport matplotlib.pyplot as pltimport numpy as np from keras.preprocessing.image import ImageDataGeneratorfrom keras import layersfrom keras.callbacks import EarlyStoppingdef Robust_CNN():        model = Sequential()    model.add(Conv2D(256, (3, 3), activation='relu', padding='same', init='glorot_uniform', input_shape=(2,128,1)))    model.add(BatchNormalization())    model.add(MaxPooling2D(pool_size=(1, 2), padding='valid',  data_format=None))    model.add(layers.Dropout(.3))    model.add(Conv2D(128, (3, 3), activation='relu', init='glorot_uniform', padding='same'))    model.add(BatchNormalization())    model.add(MaxPooling2D(pool_size=(1, 2), padding='valid', data_format=None))    model.add(layers.Dropout(.3))    model.add(Conv2D(64, (3, 3), activation='relu', init='glorot_uniform', padding='same'))    model.add(BatchNormalization())    model.add(MaxPooling2D(pool_size=(1, 2), padding='valid', data_format=None))    model.add(layers.Dropout(.3))    model.add(Conv2D(64, (3, 3), activation='relu', init='glorot_uniform', padding='same'))    model.add(BatchNormalization())    model.add(MaxPooling2D(pool_size=(1, 2), padding='valid', data_format=None))    model.add(layers.Dropout(.3))    model.add(Flatten())    model.add(Dense(128, activation='relu', init='he_normal'))    model.add(BatchNormalization())    model.add(Dense(11, activation='softmax', init='he_normal'))        return model即使我已經導入了 BatchNormalization，似乎也無法弄清楚為什么會這樣。