which version of tensorflow/keras/python is?

[JasonYang8119]

Hello, when I tried to reproduce your article, using your mmFall.ipynb. I met these bugs:
And it seems the reason is dismatch of version of tensorflow/keras/python?

My version: python=3.9, tensorflow=2.7 with keras include

So could you show your tf/keras/python version?

Thanks!

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
/tmp/ipykernel_17888/961939578.py in <module>
----> 1 HVRAE_result = AE.HVRAE_SL_predict(samples)

/tmp/ipykernel_17888/1483391771.py in HVRAE_SL_predict(self, inferencedata)
    294             # Load the saved model
    295 
--> 296         model = load_model(self.model_dir + 'HVRAE_SL_mdl.h5', compile=False,
    297                            custom_objects={'sampling': sampling_predict, 'tf': tf})
    298         adam = optimizers.Adam(lr=0.001, beta_1=0.9, beta_2=0.999, amsgrad=False)

/opt/anaconda3/envs/yjtf/lib/python3.9/site-packages/keras/saving/save.py in load_model(filepath, custom_objects, compile, options)
    198         if (h5py is not None and
    199             (isinstance(filepath, h5py.File) or h5py.is_hdf5(filepath))):
--> 200           return hdf5_format.load_model_from_hdf5(filepath, custom_objects,
    201                                                   compile)
    202 

/opt/anaconda3/envs/yjtf/lib/python3.9/site-packages/keras/saving/hdf5_format.py in load_model_from_hdf5(filepath, custom_objects, compile)
    178       model_config = model_config.decode('utf-8')
    179     model_config = json_utils.decode(model_config)
--> 180     model = model_config_lib.model_from_config(model_config,
    181                                                custom_objects=custom_objects)
    182 

/opt/anaconda3/envs/yjtf/lib/python3.9/site-packages/keras/saving/model_config.py in model_from_config(config, custom_objects)
     50                     '`Sequential.from_config(config)`?')
     51   from keras.layers import deserialize  # pylint: disable=g-import-not-at-top
---> 52   return deserialize(config, custom_objects=custom_objects)
     53 
     54 

/opt/anaconda3/envs/yjtf/lib/python3.9/site-packages/keras/layers/serialization.py in deserialize(config, custom_objects)
    206   """
    207   populate_deserializable_objects()
--> 208   return generic_utils.deserialize_keras_object(
    209       config,
    210       module_objects=LOCAL.ALL_OBJECTS,

/opt/anaconda3/envs/yjtf/lib/python3.9/site-packages/keras/utils/generic_utils.py in deserialize_keras_object(identifier, module_objects, custom_objects, printable_module_name)
    672 
    673       if 'custom_objects' in arg_spec.args:
--> 674         deserialized_obj = cls.from_config(
    675             cls_config,
    676             custom_objects=dict(

/opt/anaconda3/envs/yjtf/lib/python3.9/site-packages/keras/engine/training.py in from_config(cls, config, custom_objects)
   2395     with generic_utils.SharedObjectLoadingScope():
   2396       input_tensors, output_tensors, created_layers = (
-> 2397           functional.reconstruct_from_config(config, custom_objects))
   2398       # Initialize a model belonging to `cls`, which can be user-defined or
   2399       # `Functional`.

/opt/anaconda3/envs/yjtf/lib/python3.9/site-packages/keras/engine/functional.py in reconstruct_from_config(config, custom_objects, created_layers)
   1271   # First, we create all layers and enqueue nodes to be processed
   1272   for layer_data in config['layers']:
-> 1273     process_layer(layer_data)
   1274   # Then we process nodes in order of layer depth.
   1275   # Nodes that cannot yet be processed (if the inbound node

/opt/anaconda3/envs/yjtf/lib/python3.9/site-packages/keras/engine/functional.py in process_layer(layer_data)
   1253       from keras.layers import deserialize as deserialize_layer  # pylint: disable=g-import-not-at-top
   1254 
-> 1255       layer = deserialize_layer(layer_data, custom_objects=custom_objects)
   1256       created_layers[layer_name] = layer
   1257 

/opt/anaconda3/envs/yjtf/lib/python3.9/site-packages/keras/layers/serialization.py in deserialize(config, custom_objects)
    206   """
    207   populate_deserializable_objects()
--> 208   return generic_utils.deserialize_keras_object(
    209       config,
    210       module_objects=LOCAL.ALL_OBJECTS,

/opt/anaconda3/envs/yjtf/lib/python3.9/site-packages/keras/utils/generic_utils.py in deserialize_keras_object(identifier, module_objects, custom_objects, printable_module_name)
    672 
    673       if 'custom_objects' in arg_spec.args:
--> 674         deserialized_obj = cls.from_config(
    675             cls_config,
    676             custom_objects=dict(

/opt/anaconda3/envs/yjtf/lib/python3.9/site-packages/keras/layers/core.py in from_config(cls, config, custom_objects)
   1003   def from_config(cls, config, custom_objects=None):
   1004     config = config.copy()
-> 1005     function = cls._parse_function_from_config(
   1006         config, custom_objects, 'function', 'module', 'function_type')
   1007 

/opt/anaconda3/envs/yjtf/lib/python3.9/site-packages/keras/layers/core.py in _parse_function_from_config(cls, config, custom_objects, func_attr_name, module_attr_name, func_type_attr_name)
   1055     elif function_type == 'lambda':
   1056       # Unsafe deserialization from bytecode
-> 1057       function = generic_utils.func_load(
   1058           config[func_attr_name], globs=globs)
   1059     elif function_type == 'raw':

/opt/anaconda3/envs/yjtf/lib/python3.9/site-packages/keras/utils/generic_utils.py in func_load(code, defaults, closure, globs)
    787   except (UnicodeEncodeError, binascii.Error):
    788     raw_code = code.encode('raw_unicode_escape')
--> 789   code = marshal.loads(raw_code)
    790   if globs is None:
    791     globs = globals()

ValueError: bad marshal data (unknown type code)

[lp0617]

Do you have the solution of this problem? I get the same problem now.
I use thetensorflow-gpu=2.4.0, python=3.8

[SeanHu2100]

I recommend rewriting the model code using PyTorch to fix the version mismatch issue.