张量大小有问题，堆栈要求每个张量大小相等

Question

我正在尝试执行GAN算法，准备数据集时出错

代码如下：

 from keras.preprocessing.image import ImageDataGenerator
 import tensorflow as tf
    
    train_image_generator = ImageDataGenerator(
                                              rescale=1./255)
                                               
    dataset = train_image_generator.flow_from_directory(
                                                        batch_size=128, 
                                                        directory='../input/main-dataset110/Train',
                                                        shuffle=False, 
                                                        target_size=(28, 28), 
                                                        )

错误：

stack expects each tensor to be equal size, but got [3, 10, 40] at entry 0 and [3, 8, 43] at entry 1