Tensorflow InvalidArgumentError:不成功的TensorSliceReader构造函数：无法获取匹配的文件。。。文件系统方案'；[本地]'；未实施

我在运行笔记本时遇到以下错误：

InvalidArgumentErrorTraceback (most recent call last)
    <ipython-input-77-e16e0fd6d275> in <module>()
    ----> 1 tpu_ops = tf.contrib.tpu.batch_parallel(run_find_closest_latent_vector, [], num_shards=8)
      2 
      3 def run_once():
      4   session_tpu.run(tpu_ops)
      5 

    /usr/local/lib/python2.7/dist-packages/tensorflow/contrib/tpu/python/tpu/tpu.pyc in batch_parallel(computation, inputs, num_shards, infeed_queue, device_assignment, name)
    981       infeed_queue=infeed_queue,
    982       device_assignment=device_assignment,
    --> 983       name=name)
    984 
    985 

    /usr/local/lib/python2.7/dist-packages/tensorflow/contrib/tpu/python/tpu/tpu.pyc in shard(computation, inputs, num_shards, input_shard_axes, outputs_from_all_shards, output_shard_axes, infeed_queue, device_assignment, name)
    879       infeed_queue=infeed_queue,
    880       device_assignment=device_assignment,
    --> 881       name=name)
    882 
    883   # There must be at least one shard since num_shards > 0.

    /usr/local/lib/python2.7/dist-packages/tensorflow/contrib/tpu/python/tpu/tpu.pyc in replicate(computation, inputs, infeed_queue, device_assignment, name)
    505   """
    506   return split_compile_and_replicate(computation, inputs, infeed_queue,
    --> 507                                      device_assignment, name)[1]
    508 
    509 

    /usr/local/lib/python2.7/dist-packages/tensorflow/contrib/tpu/python/tpu/tpu.pyc in split_compile_and_replicate(***failed resolving arguments***)
    682       vscope.set_custom_getter(custom_getter)
    683 
    --> 684       outputs = computation(*computation_inputs)
    685 
    686       vscope.set_use_resource(saved_use_resource)

    <ipython-input-76-66eb3bb2ffa2> in run_find_closest_latent_vector()
     34 
     35 def run_find_closest_latent_vector():
    ---> 36   result = find_closest_latent_vector(num_optimization_steps=40)
     37   display_images(result[0], [("Loss: %.2f" % loss) for loss in result[1]])
     38 

    <ipython-input-76-66eb3bb2ffa2> in find_closest_latent_vector(num_optimization_steps)
     22 
     23     with tf.Session(tpu_address) as session_tpu:
    ---> 24       session_tpu.run(tf.global_variables_initializer())
     25       optimizer = tf.train.AdamOptimizer(learning_rate=0.3)
     26       train = optimizer.minimize(loss)

    /usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in run(self, fetches, feed_dict, options, run_metadata)
    927     try:
    928       result = self._run(None, fetches, feed_dict, options_ptr,
    --> 929                          run_metadata_ptr)
    930       if run_metadata:
    931         proto_data = tf_session.TF_GetBuffer(run_metadata_ptr)

    /usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in _run(self, handle, fetches, feed_dict, options, run_metadata)
    1150     if final_fetches or final_targets or (handle and feed_dict_tensor):
    1151       results = self._do_run(handle, final_targets, final_fetches,
    -> 1152                              feed_dict_tensor, options, run_metadata)
    1153     else:
    1154       results = []

    /usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in _do_run(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)
    1326     if handle is None:
    1327       return self._do_call(_run_fn, feeds, fetches, targets, options,
    -> 1328                            run_metadata)
    1329     else:
    1330       return self._do_call(_prun_fn, handle, feeds, fetches)

    /usr/local/lib/python2.7/dist-packages/tensorflow/python/client/session.pyc in _do_call(self, fn, *args)
    1346           pass
    1347       message = error_interpolation.interpolate(message, self._graph)
    -> 1348       raise type(e)(node_def, op, message)
    1349 
    1350   def _extend_graph(self):

    InvalidArgumentError: Unsuccessful TensorSliceReader constructor: Failed to get matching files on /tmp/tfhub_modules/2f9e2f0be115550c7ae9b90bb71b29e76fa404d8/variables/variables: Unimplemented: File system scheme '[local]' not implemented (file: '/tmp/tfhub_modules/2f9e2f0be115550c7ae9b90bb71b29e76fa404d8/variables/variables')
    [[node checkpoint_initializer_9 (defined at /usr/local/lib/python2.7/dist-packages/tensorflow_hub/native_module.py:395)  = RestoreV2[dtypes=[DT_FLOAT], _device="/job:tpu_worker/replica:0/task:0/device:CPU:0"](checkpoint_initializer/prefix, checkpoint_initializer_9/tensor_names, checkpoint_initializer/shape_and_slices)]]

    def _get_beta_accumulators(self):
        return self._beta1_power, self._beta2_power

    def find_closest_latent_vector(num_optimization_steps):
      images = []
      losses = []
      with tf.Graph().as_default():
        module = hub.Module("https://tfhub.dev/google/progan-128/1")

        initial_vector = tf.random_normal([1, latent_dim], seed=5)

        vector = tf.get_variable("vector", initializer=initial_vector)
        image = module(vector)

        target_image_difference = tf.reduce_sum(
            tf.losses.absolute_difference(image[0], target_image[:,:,:3]))

        # The latent vectors were sampled from a normal distribution. We can get
        # more realistic images if we regularize the length of the latent vector to 
        # the average length of vector from this distribution.
        regularizer = tf.abs(tf.norm(vector) - np.sqrt(latent_dim))

        loss = target_image_difference + regularizer

        with tf.Session(tpu_address) as session_tpu:
          session_tpu.run(tf.global_variables_initializer())
          optimizer = tf.train.AdamOptimizer(learning_rate=0.3)
          train = optimizer.minimize(loss)

          for _ in range(num_optimization_steps):
            _, loss_out, im_out = session_tpu.run([train, loss, image])
            print(im_out[0])
            losses.append(loss_out)
            print loss_out
        return images, losses

    def run_find_closest_latent_vector():
      result = find_closest_latent_vector(num_optimization_steps=40)
      display_images(result[0], [("Loss: %.2f" % loss) for loss in result[1]])

希望这将有助于调试该问题

---

有什么想法吗？

你能用笔记本单元格中的相关代码编辑这个问题吗？我知道你的笔记本是私人的，但也许你可以共享一个单元来帮助我们调试你的问题

我认为您的根本问题在于您使用本地机器进行输入、建模或检查点设置。但据此,

所有输入文件和模型目录必须使用云存储 bucket路径（gs://bucket name/…），并且该bucket必须是可访问的 从TPU服务器。请注意，所有数据处理和模型 检查点在TPU服务器上执行，而不是在本地计算机上执行

---

展示你的才华。很少有人会访问这个外部链接，它将在互联网瞬间消失。我删除了受密码保护的图像链接仅供参考：谢谢你修改它，我本打算自己这么做的-很抱歉没有首先这么做。我认为上面的SO链接解释了这一点，但现在我不确定如何通过存储等方式来托管此内容我已经添加了失败的代码块，请告诉我这是否有帮助。从错误日志（代码堆栈）中可以看到，这一行

-->24 session\u tpu.run（tf.global\u variables\u initializer（））

被指示为问题的根本原因。根据，该行初始化全局变量。可能您的全局变量配置不正确？我试着从你的例子中复制这个错误，但这不起作用。谢谢你的回复，你想再次访问笔记本，并尝试查看真正的代码，也许你可以给我一个或两个提示，看看真正的东西。请重新发送访问请求，我将予以批准。