这是使用数据加载器在Keras训练GAN的正确方法吗？

这样做对吗？

加载数据集:

train_ds = keras.utils.image_dataset_from_directory(
    directory='Abstract_gallery/',
    labels='inferred',
    label_mode='categorical',
    image_size=(64, 64),
    batch_size=128)

使用数据集训练GAN:

y_fake = np.zeros(batch_size)
for epoch in range(epochs):
   for batch in train_ds:
      ###### train D
      z = np.random.normal(0, 1, [batch_size,100])
      generated_images = generator(z)
      dloss = discriminator.train_on_batch(generated_images, y_fake)
      y_real = np.ones(batch.shape[0])
      dloss += discriminator.train_on_batch(batch, y_real)
      Dloss.append(dloss)
      ###### train G
      y_real = np.ones(batch_size)
      z = np.random.normal(0, 1, [batch_size,100])
      gloss = GAN.train_on_batch(z, y_real)
      Gloss.append(gloss)