TF2.0 hub Universal Sentence Encoder Multilingual Sentenepieceop not registered problem

[cbahcevan]

Have I written custom code : No
OS Platform and Distribution : Windows 10 / Google Colab
TensorFlow version (use command below):tensorflow==2.0.0
Python version:Python 3.6.9
Here is my code.

import tensorflow as tf
import tensorflow_hub as hub
import tf_sentencepiece
embedding_layer = hub.load("https://tfhub.dev/google/universal-sentence-encoder-multilingual-large/3")

Problematic output

---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/framework/ops.py in _get_op_def(self, type)
   3819     try:
-> 3820       return self._op_def_cache[type]
   3821     except KeyError:

KeyError: 'SentencepieceOp'

During handling of the above exception, another exception occurred:

NotFoundError                             Traceback (most recent call last)
8 frames
/usr/local/lib/python3.6/dist-packages/tensorflow_core/python/framework/ops.py in _get_op_def(self, type)
   3822       with c_api_util.tf_buffer() as buf:
   3823         # pylint: disable=protected-access
-> 3824         c_api.TF_GraphGetOpDef(self._c_graph, compat.as_bytes(type), buf)
   3825         # pylint: enable=protected-access
   3826         data = c_api.TF_GetBuffer(buf)

NotFoundError: Op type not registered 'SentencepieceOp' in binary running on e2f7765a82a9. Make sure the Op and Kernel are registered in the binary running in this process. Note that if you are loading a saved graph which used ops from tf.contrib, accessing (e.g.) `tf.contrib.resampler` should be done before importing the graph, as contrib ops are lazily registered when the module is first accessed

I could not get answer to this issue on tensorflow respository . I want to get multilingual sentence embeddings using pretrained Google Universal sentence encoder. But I could not able to get it from Windows 10 and Google Colab even if the internet shows diffrent. I was able to get sentence vectors using diffrent hub link like https://tfhub.dev/google/universal-sentence-encoder/4. That is not dependent to sentencepiece.

[Arkin1]

Hello, I think the multilingual module is dependent on tensorflow_text as stated in this page: https://tfhub.dev/google/universal-sentence-encoder-multilingual/3 .
I have the exact problem myself and it seems that the required library doesn't have support for windows.

[cbahcevan]

Thank you, now It is working on Google Colab. I added import tensorflow_text line. I gave up on Windows.

[thusithaC]

adding import tensorflow_text at the top works for me on linux. Thanks for the tip.

[dswah]

adding import tensorflow_text at the top works for me on linux. Thanks for the tip.

Thanks!
This worked for me too, but is pretty unsatisfying. 😒

[yousefmasry4]

This worked for me 💯

!python3.8 -m pip install tensorflow_text==2.3.0