Tensorflow hub issue: Can't load embedding model after last update

Question

This code has been working until last tf_hub update. I think, the problem is in the tensorflow_text module, that I haven't installed. But when I try to execute "pip install tensorflow_text==2.3.0" command (copied from the official tf_hub page) it throws back the error. I also tried to install it manually from github repo, but the package is still not available. Thanks.

embed = hub.load("https://tfhub.dev/google/universal-sentence-encoder/4")
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
G:\Anaconda\lib\site-packages\tensorflow\python\training\py_checkpoint_reader.py in get_tensor(self, tensor_str)
     69     return CheckpointReader.CheckpointReader_GetTensor(
---> 70         self, compat.as_bytes(tensor_str))
     71   # TODO(b/143319754): Remove the RuntimeError casting logic once we resolve the

RuntimeError: 

During handling of the above exception, another exception occurred:

OpError                                   Traceback (most recent call last)
<ipython-input-8-c716b77a9bc1> in <module>
----> 1 embed = hub.load("https://tfhub.dev/google/universal-sentence-encoder/4")

G:\Anaconda\lib\site-packages\tensorflow_hub\module_v2.py in load(handle, tags, options)
    112         module_path, tags=tags, options=options)
    113   else:
--> 114     obj = tf_v1.saved_model.load_v2(module_path, tags=tags)
    115   obj._is_hub_module_v1 = is_hub_module_v1  # pylint: disable=protected-access
    116   return obj

G:\Anaconda\lib\site-packages\tensorflow\python\saved_model\load.py in load(export_dir, tags, options)
    601     ValueError: If `tags` don't match a MetaGraph in the SavedModel.
    602   """
--> 603   return load_internal(export_dir, tags, options)
    604 
    605 

G:\Anaconda\lib\site-packages\tensorflow\python\saved_model\load.py in load_internal(export_dir, tags, options, loader_cls)
    631       try:
    632         loader = loader_cls(object_graph_proto, saved_model_proto, export_dir,
--> 633                             ckpt_options)
    634       except errors.NotFoundError as err:
    635         raise FileNotFoundError(

G:\Anaconda\lib\site-packages\tensorflow\python\saved_model\load.py in __init__(self, object_graph_proto, saved_model_proto, export_dir, ckpt_options)
    129 
    130     self._load_all()
--> 131     self._restore_checkpoint()
    132 
    133     for node in self._nodes:

G:\Anaconda\lib\site-packages\tensorflow\python\saved_model\load.py in _restore_checkpoint(self)
    328                                   self._checkpoint_options).expect_partial()
    329     else:
--> 330       load_status = saver.restore(variables_path, self._checkpoint_options)
    331     load_status.assert_existing_objects_matched()
    332     checkpoint = load_status._checkpoint

G:\Anaconda\lib\site-packages\tensorflow\python\training\tracking\util.py in restore(self, save_path, options)
   1280       dtype_map = reader.get_variable_to_dtype_map()
   1281     try:
-> 1282       object_graph_string = reader.get_tensor(base.OBJECT_GRAPH_PROTO_KEY)
   1283     except errors_impl.NotFoundError:
   1284       # The object graph proto does not exist in this checkpoint. Try the

G:\Anaconda\lib\site-packages\tensorflow\python\training\py_checkpoint_reader.py in get_tensor(self, tensor_str)
     72   # issue with throwing python exceptions from C++.
     73   except RuntimeError as e:
---> 74     error_translator(e)
     75 
     76 

G:\Anaconda\lib\site-packages\tensorflow\python\training\py_checkpoint_reader.py in error_translator(e)
     46     raise errors_impl.InternalError(None, None, error_message)
     47   else:
---> 48     raise errors_impl.OpError(None, None, error_message, errors_impl.UNKNOWN)
     49 
     50 

OpError: 

C:\Users\usr>pip install tensorflow_text v==2.3.0 ERROR: Could not find a version that satisfies the requirement tensorflow_text (from versions: none) ERROR: No matching distribution found for tensorflow_text

Answer 1

This code is working fine in Tensorflow 2.7 in Anaconda jupyter notebook. Please specify the Tensorflow version you are using while running this code.

!pip install tensorflow-hub
import tensorflow_hub as hub
embed = hub.load("https://tfhub.dev/google/universal-sentence-encoder/4")

There is no need to install the tensorflow_text to run this piece of code. However I have found there is typo error in your traceback error while installing tensorflow_text.

It should be as below:

!pip install tensorflow_text==2.3.0

Please check this link to access the Universal Sentence Encoder on TF-Hub.