CIFAR10归一化问题

初始代码：

def color_normalize(train_images, test_images):
    mean = [np.mean(train_images[:, :, :, i]) for i in range(3)]  # [125.307, 122.95, 113.865]
    std = [np.std(train_images[:, :, :, i]) for i in range(3)]  # [62.9932, 62.0887, 66.7048]
    for i in range(3):
        train_images[:, :, :, i] = (train_images[:, :, :, i] - mean[i]) / std[i]
        test_images[:, :, :, i] = (test_images[:, :, :, i] - mean[i]) / std[i]
    return train_images, test_image
 def main():
    (x_train, y_train), (x_test, y_test) = tf.keras.datasets.cifar10.load_data()
    # x代表数据，y代表标签，train代表是用于训练的，test是用于测试的
    x_train, x_test = color_normalize(x_train, x_test)  # 归一化
    labeled_data, labels = CIFAR_DataLabel(x_train, y_train)  # 对训练集进行预处理

此时，x_train, x_test只有三种数据0，1，255，明显没有达到归一化的要求，调试后发现，x_train, x_test的数据类型是uint8，直觉觉得应该改成float类型的，改成下面代码后，归一化正常

def color_normalize(train_images, test_images):
    mean = [np.mean(train_images[:, :, :, i]) for i in range(3)]  # [125.307, 122.95, 113.865]
    std = [np.std(train_images[:, :, :, i]) for i in range(3)]  # [62.9932, 62.0887, 66.7048]
    for i in range(3):
        train_images[:, :, :, i] = (train_images[:, :, :, i] - mean[i]) / std[i]
        test_images[:, :, :, i] = (test_images[:, :, :, i] - mean[i]) / std[i]
    return train_images, test_image
 def main():
    (x_train, y_train), (x_test, y_test) = tf.keras.datasets.cifar10.load_data()
    # x代表数据，y代表标签，train代表是用于训练的，test是用于测试的
    x_train = np.array(x_train, dtype=np.float64)
    x_test = np.array(x_test, dtype=np.float64)
    x_train, x_test = color_normalize(x_train, x_test)  # 归一化
    labeled_data, labels = CIFAR_DataLabel(x_train, y_train)  # 对训练集进行预处理