tensorflow/text

When downloading trained models and loading saved model, it gives `OSError: SavedModel file does not exist at: C:\Users\user\AppData\Local\Temp\tfhub_modules\602d30248ff7929470db09f7385fc895e9ceb4c0\{saved_model.pbtxt|saved_model.pb}`

HermanTam opened this issue · 5 comments

HermanTam commented

The codes were working with no such errors last month. However, I tried running the same codes today and the following codes generate the OS error. It generates the following OS Error when I tried to download the trained models (bert_preprocessor & bert_encoder) with correct URLs or load the saved model with the correct path as they worked last time. I am curious to know why that is. Thanks!

bert_preprocessor = hub.KerasLayer('https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3')

OSError                                   Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_10908/280207346.py in <module>
----> 1 bert_preprocessor = hub.KerasLayer('https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3')
      2 bert_encoder = hub.KerasLayer('https://tfhub.dev/tensorflow/bert_en_uncased_L-12_H-768_A-12/4')

~\AppData\Local\Programs\Python\Python39\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)

~\AppData\Local\Programs\Python\Python39\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 

~\AppData\Local\Programs\Python\Python39\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

~\AppData\Roaming\Python\Python39\site-packages\tensorflow\python\saved_model\load.py in load(export_dir, tags, options)
    798   if isinstance(export_dir, os.PathLike):
    799     export_dir = os.fspath(export_dir)
--> 800   result = load_partial(export_dir, None, tags, options)["root"]
    801   return result
    802 

~\AppData\Roaming\Python\Python39\site-packages\tensorflow\python\saved_model\load.py in load_partial(export_dir, filters, tags, options)
    903     tags = nest.flatten(tags)
    904   saved_model_proto, debug_info = (
--> 905       loader_impl.parse_saved_model_with_debug_info(export_dir))
    906 
    907   if (len(saved_model_proto.meta_graphs) == 1 and

~\AppData\Roaming\Python\Python39\site-packages\tensorflow\python\saved_model\loader_impl.py in parse_saved_model_with_debug_info(export_dir)
     55     parsed. Missing graph debug info file is fine.
     56   """
---> 57   saved_model = parse_saved_model(export_dir)
     58 
     59   debug_info_path = file_io.join(

~\AppData\Roaming\Python\Python39\site-packages\tensorflow\python\saved_model\loader_impl.py in parse_saved_model(export_dir)
    113       raise IOError(f"Cannot parse file {path_to_pbtxt}: {str(e)}.")
    114   else:
--> 115     raise IOError(
    116         f"SavedModel file does not exist at: {export_dir}{os.path.sep}"
    117         f"{{{constants.SAVED_MODEL_FILENAME_PBTXT}|"

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

bert_encoder = hub.KerasLayer('https://tfhub.dev/tensorflow/bert_en_uncased_L-12_H-768_A-12/4')

OSError                                   Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_10908/453047902.py in <module>
      1 #bert_preprocessor = hub.KerasLayer('https://tfhub.dev/tensorflow/bert_en_uncased_preprocess/3')
----> 2 bert_encoder = hub.KerasLayer('https://tfhub.dev/tensorflow/bert_en_uncased_L-12_H-768_A-12/4')

~\AppData\Local\Programs\Python\Python39\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)

~\AppData\Local\Programs\Python\Python39\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 

~\AppData\Local\Programs\Python\Python39\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

~\AppData\Roaming\Python\Python39\site-packages\tensorflow\python\saved_model\load.py in load(export_dir, tags, options)
    798   if isinstance(export_dir, os.PathLike):
    799     export_dir = os.fspath(export_dir)
--> 800   result = load_partial(export_dir, None, tags, options)["root"]
    801   return result
    802 

~\AppData\Roaming\Python\Python39\site-packages\tensorflow\python\saved_model\load.py in load_partial(export_dir, filters, tags, options)
    903     tags = nest.flatten(tags)
    904   saved_model_proto, debug_info = (
--> 905       loader_impl.parse_saved_model_with_debug_info(export_dir))
    906 
    907   if (len(saved_model_proto.meta_graphs) == 1 and

~\AppData\Roaming\Python\Python39\site-packages\tensorflow\python\saved_model\loader_impl.py in parse_saved_model_with_debug_info(export_dir)
     55     parsed. Missing graph debug info file is fine.
     56   """
---> 57   saved_model = parse_saved_model(export_dir)
     58 
     59   debug_info_path = file_io.join(

~\AppData\Roaming\Python\Python39\site-packages\tensorflow\python\saved_model\loader_impl.py in parse_saved_model(export_dir)
    113       raise IOError(f"Cannot parse file {path_to_pbtxt}: {str(e)}.")
    114   else:
--> 115     raise IOError(
    116         f"SavedModel file does not exist at: {export_dir}{os.path.sep}"
    117         f"{{{constants.SAVED_MODEL_FILENAME_PBTXT}|"

OSError: SavedModel file does not exist at: C:\Users\user\AppData\Local\Temp\tfhub_modules\d760773f85f64fc84ae0b47310f7cfe3bcec4868\{saved_model.pbtxt|saved_model.pb}


loaded_model = tf.keras.models.load_model(
       ('BertForSpam.h5'),
       custom_objects={'KerasLayer':hub.KerasLayer}
)

OSError                                   Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_10908/1893849820.py in <module>
      2 #ValueError: Unknown layer: KerasLayer. Please ensure this object is passed to the `custom_objects` argument.
      3 
----> 4 loaded_model = tf.keras.models.load_model(
      5        ('BertForSpam.h5'),
      6        custom_objects={'KerasLayer':hub.KerasLayer}

~\AppData\Local\Programs\Python\Python39\lib\site-packages\keras\utils\traceback_utils.py in error_handler(*args, **kwargs)
     68             # To get the full stack trace, call:
     69             # `tf.debugging.disable_traceback_filtering()`
---> 70             raise e.with_traceback(filtered_tb) from None
     71         finally:
     72             del filtered_tb

~\AppData\Local\Programs\Python\Python39\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)

~\AppData\Local\Programs\Python\Python39\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 

~\AppData\Local\Programs\Python\Python39\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

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

bert_preprocessor
bert_preprocessor

bert_encoder
bert_encoder

load_model
load_model

broken commented

It's saying it cannot find the model file:
OSError: SavedModel file does not exist at: C:\Users\user\AppData\Local\Temp\tfhub_modules\d760773f85f64fc84ae0b47310f7cfe3bcec4868\{saved_model.pbtxt|saved_model.pb}

Can you look in that directory to see if it exists? Maybe it's saved under a different name? Or, did the operations to create it fail?

HermanTam commented

Thank you for your response.

There are two empty folders in the directory (please refer to the screenshot). I wonder why this happens because there is currently even a problem with downloading the trained models (bert_preprocessor & bert_encoder) and there was no problem with the same code last month.
image

Seems like this person below has experienced something similar:
image

HermanTam commented

I wonder why that is. I tried running the same code today and it works without the error just like the first time I ran it. I checked the folder again and there are the models.
image
image
image

broken commented

I'm glad you got it working. One thing I noticed is that these file are in your .../Local/Temp/... directory. The operating system will delete files in here after some amount of time.

My guess at a timeline of events:

  1. run code and it saves the model correctly in the directory
  2. time passes and the OS deletes the model files
  3. run code again and it finds the directory for the model so doesn't save to overwrite it even though files in the directory don't exist - thus it fails trying to load it.
  4. more time passes and the directory is now removed also
  5. run code again, directory doesn't exist, so model is correctly saved like the first time you ran it.