Question

我在Tensorflow输入管道/数据生成器中遇到Exception的问题。我自己的脚本tfdata_generator(...)中的生成器函数generator.py是通过fit_generator()中的Keras main.py方法调用的。生成器函数如下：

def tfdata_generator(...):

    image_tensor = tf.data.TFRecordDataset(path_list)  
    height_tensor = tf.data.Dataset.from_tensor_slices(height_list)  
    label_tensor = tf.data.Dataset.from_tensor_slices(label_list) 

    dataset = dataset.map(map_func=lambda a, b: ...)  # do several mappings 
    dataset = dataset.apply(tf.contrib.data.unbatch())  
    if is_training == 'True':
        dataset = dataset.shuffle(buffer_size=count)

    dataset = dataset.repeat()  
    dataset_batched = dataset.batch(batch_size=batch_size)  
    iterator = dataset_batched.make_one_shot_iterator()
    next_batch = iterator.get_next()  

    with tf.Session() as sess:
        while True:
            try:
                images, height, labels   = sess.run(next_batch)
                yield ({'input_images': images, 'input_height': height}), labels

            except IndexError:  # tf.errors.OutOfRangeError:
                print('Finished training')
                break

我在控制台中收到以下错误消息：


90/90 [==============================] - 63s 697ms/step - loss: 0.0171 - acc: 0.9889 - val_loss: 0.0026 - val_acc: 1.0000
Network Training Finished!
Saved model
Exception ignored in: <generator object tfdata_generator at 0x7ff7580f56d0>
Traceback (most recent call last):
  File "/util/generator.py", line 233, in tfdata_generator
    break
  File "/usr/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1587, in __exit__
    self._default_graph_context_manager.__exit__(exec_type, exec_value, exec_tb)
  File "/usr/lib64/python3.6/contextlib.py", line 99, in __exit__
    self.gen.throw(type, value, traceback)
  File "/usr/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 5233, in get_controller
    context.context().context_switches.pop()
  File "/usr/lib/python3.6/site-packages/tensorflow/python/eager/context.py", line 141, in pop
    self.stack.pop()
IndexError: pop from empty list
Exception ignored in: <generator object tfdata_generator at 0x7ff7580f5d58>
Traceback (most recent call last):
  File "/util/generator.py", line 233, in tfdata_generator
    break
  File "/usr/lib/python3.6/site-packages/tensorflow/python/client/session.py", line 1587, in __exit__
    self._default_graph_context_manager.__exit__(exec_type, exec_value, exec_tb)
  File "/usr/lib64/python3.6/contextlib.py", line 99, in __exit__
    self.gen.throw(type, value, traceback)
  File "/usr/lib/python3.6/site-packages/tensorflow/python/framework/ops.py", line 5233, in get_controller
    context.context().context_switches.pop()
  File "/usr/lib/python3.6/site-packages/tensorflow/python/eager/context.py", line 141, in pop
    self.stack.pop()
IndexError: pop from empty list
Process finished with exit code 0

在fit_generator()方法之后，“ Network Training结束”和“ Saveed Moedl”被打印在主屏幕上。

问题： 显然，这是发电机退出后发生的典型例外。因此，异常也将被忽略。在这种情况下，我的程序将成功完成，因此无需在Python控制台中显示被忽略的异常。简直就是丑陋。如何摆脱被忽略的异常的打印输出？

感谢您的帮助克里斯托夫

Tensorflow数据生成器/输入管道IndexError异常

0 个答案: