Tensorflow 将SessionRunHook与TPU一起使用

我已经开发了一个SessionRunHook，它连接到TPUEstimator。SessionRunHook在CPU上工作得非常好，但是如果我使用TPU，我会得到一个错误：

INFO:tensorflow:Error recorded from outfeed: Attempted to use a closed Session.
INFO:tensorflow:Error recorded from evaluation_loop: Operation 'mean_1' has been marked as not fetchable.
INFO:tensorflow:evaluation_loop marked as finished
WARNING:tensorflow:Reraising captured error
Traceback (most recent call last):
  File "/usr/lib/python3.5/runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "/usr/lib/python3.5/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "/home/donatas_repecka/workspace/bert/run_pretraining.py", line 549, in <module>
    tf.app.run()
INFO:tensorflow:Error recorded from infeed: Step was cancelled by an explicit call to `Session::Close()`.
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/platform/app.py", line 125, in run
    _sys.exit(main(argv))
  File "/home/donatas_repecka/workspace/bert/run_pretraining.py", line 505, in main
    result = estimator.evaluate(input_fn=input_fn, steps=FLAGS.max_eval_steps)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/contrib/tpu/python/tpu/tpu_estimator.py", line 2424, in evaluate
    rendezvous.raise_errors()
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/contrib/tpu/python/tpu/error_handling.py", line 128, in raise_errors
    six.reraise(typ, value, traceback)
  File "/usr/local/lib/python3.5/dist-packages/six.py", line 693, in reraise
    raise value
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/contrib/tpu/python/tpu/error_handling.py", line 101, in catch_errors
    yield
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/contrib/tpu/python/tpu/tpu_estimator.py", line 451, in _run_outfeed
    session.run(self._dequeue_ops)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py", line 929, in run
    run_metadata_ptr)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py", line 1075, in _run
    raise RuntimeError('Attempted to use a closed Session.')
RuntimeError: Attempted to use a closed Session.

INFO:tensorflow:从出料口记录的错误：试图使用关闭的会话。
信息：tensorflow:评估循环记录的错误：操作“mean_1”已标记为不可获取。
信息：tensorflow:评估循环标记为已完成
警告：tensorflow:重新释放捕获的错误
回溯（最近一次呼叫最后一次）：
文件“/usr/lib/python3.5/runpy.py”，第193行，在“运行”模块中作为“主”
“\uuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuu
文件“/usr/lib/python3.5/runpy.py”，第85行，在运行代码中
exec（代码、运行\全局）
文件“/home/donatas_repecka/workspace/bert/run_pretraining.py”，第549行，在
tf.app.run（）
信息：tensorflow:从馈入记录的错误：步骤被对“Session:：Close（）”的显式调用取消。
文件“/usr/local/lib/python3.5/dist-packages/tensorflow/python/platform/app.py”，第125行，运行中
_系统出口（主（argv））
文件“/home/donatas_repecka/workspace/bert/run_pretraining.py”，第505行，在main中
结果=估计器。评估（输入=输入，步骤=标志。最大评估步骤）
文件“/usr/local/lib/python3.5/dist packages/tensorflow/contrib/tpu/python/tpu/tpu_estimator.py”，第2424行，在evaluate中
集合。引发错误（）
文件“/usr/local/lib/python3.5/dist packages/tensorflow/contrib/tpu/python/tpu/error\u handling.py”，raise\u errors中的第128行
六、重放（类型、值、回溯）
文件“/usr/local/lib/python3.5/dist-packages/six.py”，第693行，在reraise中
增值
文件“/usr/local/lib/python3.5/dist packages/tensorflow/contrib/tpu/python/tpu/error\u handling.py”，第101行，在catch\u errors中
产量
文件“/usr/local/lib/python3.5/dist packages/tensorflow/contrib/tpu/python/tpu/tpu\u estimator.py”，第451行，in\u run\u outfeed
session.run（self.\u dequeue\u ops）
文件“/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py”，第929行，正在运行
运行_元数据_ptr）
文件“/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py”，第1075行，正在运行
raise RUNTIMERROR（'试图使用关闭的会话'）
RuntimeError:试图使用已关闭的会话。

---

还有其他人遇到过这个问题并找到了解决方法吗？

你能找到解决方法吗？不幸的是，没有，我必须重构代码