Saved Model doesn't exist error, when trying to use TensorFlow Hub

[IavTavares]

System information
-Windows 10

• TensorFlow version 2.6
• Python version 3.8
• Installed using "pip install tensorflow"

Describe the problem
I'm trying to follow tensorflow's tutorial on how to classify text with the BERT model.
However, using their notebook, I encounter a error message telling me that the model I saved does not exist.

Provide the exact sequence of commands / steps that you executed before running into the problem

I run the notebook I downloaded above. Everything goes well, until I run the following:
bert_preprocess_model = hub.KerasLayer(tfhub_handle_preprocess)

Any other info / logs

OSError Traceback (most recent call last)
in
----> 1 bert_preprocess_model = hub.KerasLayer(tfhub_handle_preprocess)

~\anaconda3\lib\site-packages\tensorflow_hub\keras_layer.py in init(self, handle, trainable, arguments, _sentinel, tags, signature, signature_outputs_as_dict, output_key, output_shape, load_options, **kwargs)
151
152 self._load_options = load_options
--> 153 self._func = load_module(handle, tags, self._load_options)
154 self._has_training_argument = func_has_training_argument(self._func)
155 self._is_hub_module_v1 = getattr(self._func, "_is_hub_module_v1", False)

~\anaconda3\lib\site-packages\tensorflow_hub\keras_layer.py in load_module(handle, tags, load_options)
447 except ImportError: # Expected before TF2.4.
448 set_load_options = load_options
--> 449 return module_v2.load(handle, tags=tags, options=set_load_options)
450
451

~\anaconda3\lib\site-packages\tensorflow_hub\module_v2.py in load(handle, tags, options)
104 module_path, tags=tags, options=options)
105 else:
--> 106 obj = tf.compat.v1.saved_model.load_v2(module_path, tags=tags)
107 obj._is_hub_module_v1 = is_hub_module_v1 # pylint: disable=protected-access
108 return obj

~\anaconda3\lib\site-packages\tensorflow\python\saved_model\load.py in load(export_dir, tags, options)
862 """
863 metrics.IncrementReadApi(_LOAD_V2_LABEL)
--> 864 result = load_internal(export_dir, tags, options)["root"]
865 metrics.IncrementRead()
866 return result

~\anaconda3\lib\site-packages\tensorflow\python\saved_model\load.py in load_internal(export_dir, tags, options, loader_cls, filters)
876 tags = nest.flatten(tags)
877 saved_model_proto, debug_info = (
--> 878 loader_impl.parse_saved_model_with_debug_info(export_dir))
879
880 if (len(saved_model_proto.meta_graphs) == 1 and

~\anaconda3\lib\site-packages\tensorflow\python\saved_model\loader_impl.py in parse_saved_model_with_debug_info(export_dir)
58 parsed. Missing graph debug info file is fine.
59 """
---> 60 saved_model = _parse_saved_model(export_dir)
61
62 debug_info_path = os.path.join(

~\anaconda3\lib\site-packages\tensorflow\python\saved_model\loader_impl.py in parse_saved_model(export_dir)
116 raise IOError("Cannot parse file %s: %s." % (path_to_pbtxt, str(e)))
117 else:
--> 118 raise IOError(
119 "SavedModel file does not exist at: %s%s{%s|%s}" %
120 (export_dir, os.path.sep, constants.SAVED_MODEL_FILENAME_PBTXT,

OSError: SavedModel file does not exist at: C:\Users\iavta\AppData\Local\Temp\tfhub_modules\602d30248ff7929470db09f7385fc895e9ceb4c0{saved_model.pbtxt|saved_model.pb}

---

---

In the directory C:\Users\iavta\AppData\Local\Temp\tfhub_modules\602d30248ff7929470db09f7385fc895e9ceb4c0

I have only two empty folders called "assets" and "variables".

[IavTavares]

Also, the problem doesn't seem to be from TensorFlow-Hub, since when I run

import tensorflow_hub as hub
model = hub.KerasLayer("https://tfhub.dev/google/nnlm-en-dim128/2")
embeddings = model(["The rain in Spain.", "falls",
                      "mainly", "In the plain!"])

print(embeddings.shape)  #(4,128)

Everything runs smoothly, and I get the desired output.

[sushreebarsa]

@IavTavares Could you please refer to the similar issues link1, link2 ,Please have a look at the common issues and let us know if it helps?Thank you!

[IavTavares]

I tried one of the answers in link1.

I downloaded the preprocessing model, and redirected the hub.load() statement to the model's directory.

I also had to do something similar for the encoder.

[sushreebarsa]

@IavTavares Thank you for the update! Could you please let us know if this issue is resolved ? Thanks!

[IavTavares]

Well... it depends on what you mean by resolved. If you mean did I find a workaround? Yes. If the underlying issue was solved? No. Sent from Mail<https://go.microsoft.com/fwlink/?LinkId=550986> for Windows From: ***@***.***> Sent: 22 October 2021 04:49 To: ***@***.***> Cc: ***@***.***>; ***@***.***> Subject: Re: [tensorflow/tensorflow] Saved Model doesn't exist error, when trying to use TensorFlow Hub (Issue tensorflow/tensorflow#52606) @IavTavares<https://github.com/IavTavares> Thank you for the update! Could you please let us know if this issue is resolved ? Thanks! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<https://github.com/tensorflow/tensorflow/issues/52606#issuecomment-949262894>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AV4NS6PS7RCEXJ6FSSVQHATUIDNLJANCNFSM5GNNZ5ZQ>. Triage notifications on the go with GitHub Mobile for iOS<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

[sanatmpa1]

@IavTavares,

Are you getting this error only when loading the model small_bert/bert_en_uncased_L-4_H-512_A-8? Can you try using ert_model_name = small_bert/bert_en_uncased_L-2_H-128_A-2 and let us know if you're getting the same error for that one as well? Thanks!

[IavTavares]

If I run tfhub_handle_preprocess="https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3" bert_preprocess_model = hub.KerasLayer(tfhub_handle_preprocess) tfhub_handle_encoder="https://tfhub.dev/tensorflow/small_bert/bert_en_uncased_L-12_H-128_A-2/2" bert_model = hub.KerasLayer(tfhub_handle_encoder) Then I get the same error. However, if I point tfhub_handle_preprocess to a directory with the preprocess already downloaded, then everything works fine. Sent from Mail<https://go.microsoft.com/fwlink/?LinkId=550986> for Windows From: ***@***.***> Sent: 22 October 2021 16:22 To: ***@***.***> Cc: ***@***.***>; ***@***.***> Subject: Re: [tensorflow/hub] Saved Model doesn't exist error, when trying to use TensorFlow Hub (Issue #816) @IavTavares<https://github.com/IavTavares>, Are you getting this error only when loading the model small_bert/bert_en_uncased_L-4_H-512_A-8? Can you try using ert_model_name = small_bert/bert_en_uncased_L-2_H-128_A-2 and let us know if you're getting the same error for that one as well? Thanks! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub<#816 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AV4NS6JYBDWEQKGNQ5EEAT3UIF6R5ANCNFSM5GQVNJ2A>. Triage notifications on the go with GitHub Mobile for iOS<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

[sanatmpa1]

@IavTavares

Can you try providing a custom directory for TFHUB_CACHE_DIR and let us know if the issue still persists? You can refer this link

[UsharaniPagadala]

@IavTavares

Closing this issue due to inactivity.Please feel free to reopen if this issue still persists.Thanks