IndexError：我的Keras模型中的元组索引超出范围。

Question

我对Keras / Tensorflow完全陌生。下面是我的fit_generator

train_dataset=train_fn_inputs(batch_size, None) 
val_data=validation_fn_inputs(batch_size, None) 
total_records = 44712  
val_records = 11178  
steps_per_epoch=int(total_records // batch_size)

hist=model.fit_generator(#aug.flow(X_def, y_def, batch_size=batch_size), 
               #get_batches(X_def, y_def, batch_size), 
               train_dataset,
               steps_per_epoch=steps_per_epoch, #(training_df.shape[0])//batchsize,
               epochs=5,
               verbose = 1,
               #callbacks=[early_stopping],
               #validation_data=val_data, 
               #validation_steps=val_records//batch_size,
               workers=0
          )

函数定义为：

def train_fn_inputs(bs,  aug=None):
    train_files, total_records = get_training_data_old()
    steps_per_epoch = int(total_records / batch_size)    
    raw_dataset = tf.data.TFRecordDataset(train_files)     #.repeat()
    parsed_image_dataset = raw_dataset.map(_parse_image_function).repeat().shuffle(buffer_size=buf_size).batch(batch_size).make_initializable_iterator()  

    image, label = parsed_image_dataset.get_next()
    image = tf.reshape(image, [3, IMG_WIDTH, IMG_HEIGHT, bs])
    #label = tf.reshape(label, [bs, 75, 25])

    while True:
       yield (np.array(image), np.array(label))

但是出现此错误：

  

文件“ ... \ Anaconda3 \ lib \ site-packages \ tensorflow \ python \ keras \ engine \ training_generator.py”，行184，在model_iteration中       batch_size = int（nest.flatten（batch_data）[0] .shape [0]）

     

IndexError：元组索引超出范围