Keras:从自定义生成器中获取一个批次

我用这里的代码创建了一个自定义生成器 https://github.com/keras-team/keras/issues/3386#issuecomment-237555199 作为模板。而不是使用 .flow .flow_from_dataframe . 因此,我的发电机看起来是这样的:

def createGenerator( X, I):

    while True:

        # suffled indices    
        idx = np.random.permutation( X.shape[0])
        # create image generator
        datagen=ImageDataGenerator(preprocessing_function=preprocess_input,
                              horizontal_flip = True,
                              rotation_range=20)


        batches = datagen.flow_from_dataframe(shuffle=False,dataframe=X[["url",target]], directory=bilder_resized_dir, x_col="url", y_col=target,
                                            has_ext=False, class_mode="categorical", color_mode=color_mode, target_size=(IMG_SIZE,IMG_SIZE), batch_size=batch_size,classes=class_list)
        idx0 = 0
        for batch in batches:
            idx1 = idx0 + batch[0].shape[0]

            yield [batch[0], I[ idx[ idx0:idx1 ] ]], batch[1]

            idx0 = idx1
            if idx1 >= X.shape[0]:
                break

next() 就像这样:

combo_gen = createGenerator(X1_dataframe ,X2_aux_input)
x,y = next(combo_gen)

但是,这会导致以下错误:

---------------------------------------------------------------------------
StopIteration                             Traceback (most recent call last)
<ipython-input-274-4a67da5d38d9> in <module>
----> 1 x,y = next(combo_gen)

StopIteration: 

我怎样才能从这台发电机上得到一批呢?