Models: Saving (& Loading) models

Created on 14 Jul 2020 · 10Comments · Source: tensorflow/models

TF_2_2_colab_object_detection_20200713b_TF2_2_cut_down.zip

Prerequisites

Please answer the following questions for yourself before submitting an issue.

[ ] I am using the latest TensorFlow Model Garden release and TensorFlow 2.
[ ] I am reporting the issue to the correct repository. (Model Garden official or research directory)
[x] I checked to make sure that this issue has not been filed already.

1. The entire URL of the file you are using

https://github.com/tensorflow/models/blob/master/research/object_detection/colab_tutorials/eager_few_shot_od_training_tf2_colab.ipynb

Please see attached zip file with jupyter notebook based on the 'eager_few_shot_od_training_tf2_colab.ipynb' authored by 'Tombstone' at google. This is currently inaccessible at time of posting, but is listed under:
https://github.com/tensorflow/models/tree/master/research/object_detection/colab_tutorials

Differences:

no import or use of colab_utils
- my data substituted for rubber duck image set
  
  -(training on my set has been fine)
- import of xml data direct from xml files to 'gt_boxes' ariable for training
- use of 800x800 pixel images (in place of 640x640)
added model.save and tf.saved_model.save() (& as yet untried model.load). : These may be inappropriate for object detection models?
- test set functionality illustrated with images from training set (as provided)
- nb: please substitute director paths and filenames for your own systems...

Migrated to Jupyter notebook in newly created Tensorflow 2.2 environment

Apologies: Should be under 'RESEARCHmodels'

2. Describe the bug

A clear and concise description of what the bug is:
Model trains successfully but both
1) model.save
in form
detection_model.save(model_dest)

2) tf.saved_mode.save()
in form
tf.saved_model.save(
detection_model, model_dest, signatures=None, options=None
)
both yield errors

3. Steps to reproduce

Steps to reproduce the behavior.

# save trained model : model.save()
model_directory = '/home/michael/jupyter_notebooks_TF_2_2/models/'
model_name = 'TF_2_2_colab_DOT_object_detection_20200713a'
model_dest = os.path.join(os.sep, model_directory, model_name)
detection_model.save(model_dest)

yields output:

AttributeError Traceback (most recent call last)
in
9 #)
10
---> 11 detection_model.save(model_dest)

AttributeError: 'SSDMetaArch' object has no attribute 'save'

while
``# saved_model trained model

model_directory = '/home/michael/jupyter_notebooks_TF_2_2/models/'
model_name = 'TF_2_2_colab_DOT_object_detection_20200713a'
model_dest = os.path.join(os.sep, model_directory, model_name)

tf.saved_model.save(to_export, '/tmp/adder')

tf.saved_model.save(
detection_model, model_dest, signatures=None, options=None
)

WARNING:tensorflow:Skipping full serialization of Keras layer , because it is not built.

TypeError Traceback (most recent call last)
in
6 #tf.saved_model.save(to_export, '/tmp/adder')
7
----> 8 tf.saved_model.save(
9 detection_model, model_dest, signatures=None, options=None
10 )

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/saved_model/save.py in save(obj, export_dir, signatures, options)
948 meta_graph_def = saved_model.meta_graphs.add()
949
--> 950 _, exported_graph, object_saver, asset_info = _build_meta_graph(
951 obj, export_dir, signatures, options, meta_graph_def)
952 saved_model.saved_model_schema_version = constants.SAVED_MODEL_SCHEMA_VERSION

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/saved_model/save.py in _build_meta_graph(obj, export_dir, signatures, options, meta_graph_def)
1020 # Note we run this twice since, while constructing the view the first time
1021 # there can be side effects of creating variables.
-> 1022 _ = _SaveableView(checkpoint_graph_view)
1023 saveable_view = _SaveableView(checkpoint_graph_view, wrapped_functions)
1024 object_saver = util.TrackableSaver(checkpoint_graph_view)

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/saved_model/save.py in __init__(self, checkpoint_view, wrapped_functions)
171 self.checkpoint_view = checkpoint_view
172 trackable_objects, node_ids, slot_variables = (
--> 173 self.checkpoint_view.objects_ids_and_slot_variables())
174 self.nodes = trackable_objects
175 self.node_ids = node_ids

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/training/tracking/graph_view.py in objects_ids_and_slot_variables(self)
413 A tuple of (trackable objects, object -> node id, slot variables)
414 """
--> 415 trackable_objects, path_to_root = self._breadth_first_traversal()
416 object_names = object_identity.ObjectIdentityDictionary()
417 for obj, path in path_to_root.items():

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/training/tracking/graph_view.py in _breadth_first_traversal(self)
197 % (current_trackable,))
198 bfs_sorted.append(current_trackable)
--> 199 for name, dependency in self.list_dependencies(current_trackable):
200 if dependency not in path_to_root:
201 path_to_root[dependency] = (

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/saved_model/save.py in list_dependencies(self, obj)
106 def list_dependencies(self, obj):
107 """Overrides a parent method to include add_object objects."""
--> 108 extra_dependencies = self.list_extra_dependencies(obj)
109 extra_dependencies.update(self._extra_dependencies.get(obj, {}))
110

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/saved_model/save.py in list_extra_dependencies(self, obj)
134
135 def list_extra_dependencies(self, obj):
--> 136 return obj._list_extra_dependencies_for_serialization( # pylint: disable=protected-access
137 self._serialization_cache)
138

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/engine/base_layer.py in _list_extra_dependencies_for_serialization(self, serialization_cache)
2743
2744 def _list_extra_dependencies_for_serialization(self, serialization_cache):
-> 2745 return (self._trackable_saved_model_saver
2746 .list_extra_dependencies_for_serialization(serialization_cache))
2747

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/base_serialization.py in list_extra_dependencies_for_serialization(self, serialization_cache)
72 of attributes are listed in the saved_model._LayerAttributes class.
73 """
---> 74 return self.objects_to_serialize(serialization_cache)
75
76 def list_functions_for_serialization(self, serialization_cache):

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/layer_serialization.py in objects_to_serialize(self, serialization_cache)
70
71 def objects_to_serialize(self, serialization_cache):
---> 72 return (self._get_serialized_attributes(
73 serialization_cache).objects_to_serialize)
74

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/layer_serialization.py in _get_serialized_attributes(self, serialization_cache)
89 return serialized_attr
90
---> 91 object_dict, function_dict = self._get_serialized_attributes_internal(
92 serialization_cache)
93

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/model_serialization.py in _get_serialized_attributes_internal(self, serialization_cache)
50 # the ones serialized by Layer.
51 objects, functions = (
---> 52 super(ModelSavedModelSaver, self)._get_serialized_attributes_internal(
53 serialization_cache))
54 functions['_default_save_signature'] = default_signature

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/layer_serialization.py in _get_serialized_attributes_internal(self, serialization_cache)
99 """Returns dictionary of serialized attributes."""
100 objects = save_impl.wrap_layer_objects(self.obj, serialization_cache)
--> 101 functions = save_impl.wrap_layer_functions(self.obj, serialization_cache)
102 # Attribute validator requires that the default save signature is added to
103 # function dict, even if the value is None.

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/save_impl.py in wrap_layer_functions(layer, serialization_cache)
159 # call with losses) are traced with the same inputs.
160 call_collection = LayerCallCollection(layer)
--> 161 call_fn_with_losses = call_collection.add_function(
162 _wrap_call_and_conditional_losses(layer),
163 '{}_layer_call_and_return_conditional_losses'.format(layer.name))

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/save_impl.py in add_function(self, call_fn, name)
501 # Manually add traces for layers that have keyword arguments and have
502 # a fully defined input signature.
--> 503 self.add_trace(*self._input_signature)
504 return fn
505

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/save_impl.py in add_trace(self, args, *kwargs)
416 fn.get_concrete_function(args, *kwargs)
417
--> 418 trace_with_training(True)
419 trace_with_training(False)
420 else:

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/save_impl.py in trace_with_training(value, fn)
414 utils.set_training_arg(value, self._training_arg_index, args, kwargs)
415 with K.learning_phase_scope(value):
--> 416 fn.get_concrete_function(args, *kwargs)
417
418 trace_with_training(True)

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/save_impl.py in get_concrete_function(self, args, *kwargs)
545 if not self.call_collection.tracing:
546 self.call_collection.add_trace(args, *kwargs)
--> 547 return super(LayerCall, self).get_concrete_function(args, *kwargs)
548
549

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/eager/def_function.py in get_concrete_function(self, args, *kwargs)
957 ValueError: if this object has not yet been called on concrete values.
958 """
--> 959 concrete = self._get_concrete_function_garbage_collected(args, *kwargs)
960 concrete._garbage_collector.release() # pylint: disable=protected-access
961 return concrete

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/eager/def_function.py in _get_concrete_function_garbage_collected(self, args, *kwargs)
863 if self._stateful_fn is None:
864 initializers = []
--> 865 self._initialize(args, kwargs, add_initializers_to=initializers)
866 self._initialize_uninitialized_variables(initializers)
867

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/eager/def_function.py in _initialize(self, args, kwds, add_initializers_to)
503 self._graph_deleter = FunctionDeleter(self._lifted_initializer_graph)
504 self._concrete_stateful_fn = (
--> 505 self._stateful_fn._get_concrete_function_internal_garbage_collected( # pylint: disable=protected-access
506 args, *kwds))
507

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/eager/function.py in _get_concrete_function_internal_garbage_collected(self, args, *kwargs)
2444 args, kwargs = None, None
2445 with self._lock:
-> 2446 graph_function, _, _ = self._maybe_define_function(args, kwargs)
2447 return graph_function
2448

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/eager/function.py in _maybe_define_function(self, args, kwargs)
2775
2776 self._function_cache.missed.add(call_context_key)
-> 2777 graph_function = self._create_graph_function(args, kwargs)
2778 self._function_cache.primary[cache_key] = graph_function
2779 return graph_function, args, kwargs

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/eager/function.py in _create_graph_function(self, args, kwargs, override_flat_arg_shapes)
2655 arg_names = base_arg_names + missing_arg_names
2656 graph_function = ConcreteFunction(
-> 2657 func_graph_module.func_graph_from_py_func(
2658 self._name,
2659 self._python_function,

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/framework/func_graph.py in func_graph_from_py_func(name, python_func, args, kwargs, signature, func_graph, autograph, autograph_options, add_control_dependencies, arg_names, op_return_value, collections, capture_by_value, override_flat_arg_shapes)
979 _, original_func = tf_decorator.unwrap(python_func)
980
--> 981 func_outputs = python_func(func_args, *func_kwargs)
982
983 # invariant: func_outputs contains only Tensors, CompositeTensors,

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/eager/def_function.py in wrapped_fn(args, *kwds)
439 # __wrapped__ allows AutoGraph to swap in a converted function. We give
440 # the function a weak reference to itself to avoid a reference cycle.
--> 441 return weak_wrapped_fn().__wrapped__(args, *kwds)
442 weak_wrapped_fn = weakref.ref(wrapped_fn)
443

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/save_impl.py in wrapper(args, *kwargs)
522 saving=True):
523 with base_layer_utils.autocast_context_manager(layer._compute_dtype): # pylint: disable=protected-access
--> 524 ret = method(args, *kwargs)
525 _restore_layer_losses(original_losses)
526 return ret

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/utils.py in wrap_with_training_arg(args, *kwargs)
165 return wrapped_call(args, *kwargs)
166
--> 167 return tf_utils.smart_cond(
168 training,
169 lambda: replace_training_and_call(True),

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/utils/tf_utils.py in smart_cond(pred, true_fn, false_fn, name)
62 return control_flow_ops.cond(
63 pred, true_fn=true_fn, false_fn=false_fn, name=name)
---> 64 return smart_module.smart_cond(
65 pred, true_fn=true_fn, false_fn=false_fn, name=name)
66

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/framework/smart_cond.py in smart_cond(pred, true_fn, false_fn, name)
52 if pred_value is not None:
53 if pred_value:
---> 54 return true_fn()
55 else:
56 return false_fn()

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/utils.py in ()
167 return tf_utils.smart_cond(
168 training,
--> 169 lambda: replace_training_and_call(True),
170 lambda: replace_training_and_call(False))
171

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/utils.py in replace_training_and_call(training)
163 def replace_training_and_call(training):
164 set_training_arg(training, training_arg_index, args, kwargs)
--> 165 return wrapped_call(args, *kwargs)
166
167 return tf_utils.smart_cond(

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/save_impl.py in call_and_return_conditional_losses(inputs, args, *kwargs)
564 layer_call = _get_layer_call_method(layer)
565 def call_and_return_conditional_losses(inputs, args, *kwargs):
--> 566 return layer_call(inputs, args, *kwargs), layer.get_losses_for(inputs)
567 return _create_call_fn_decorator(layer, call_and_return_conditional_losses)
568

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/object_detection/meta_architectures/ssd_meta_arch.py in call(self, inputs, *kwargs)
249 # method.
250 def call(self, inputs, *kwargs):
--> 251 return self._extract_features(inputs)
252
253

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/object_detection/models/ssd_resnet_v1_fpn_keras_feature_extractor.py in _extract_features(self, preprocessed_inputs)
233 (feature_block, feature_block_map[feature_block])
234 for feature_block in feature_block_list]
--> 235 fpn_features = self._fpn_features_generator(fpn_input_image_features)
236
237 feature_maps = []

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/engine/base_layer.py in __call__(self, args, *kwargs)
966 with base_layer_utils.autocast_context_manager(
967 self._compute_dtype):
--> 968 outputs = self.call(cast_inputs, args, *kwargs)
969 self._handle_activity_regularization(inputs, outputs)
970 self._set_mask_metadata(inputs, outputs, input_masks)

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/utils.py in return_outputs_and_add_losses(args, *kwargs)
69 inputs = args[inputs_arg_index]
70 args = args[inputs_arg_index + 1:]
---> 71 outputs, losses = fn(inputs, args, *kwargs)
72 layer.add_loss(losses, inputs)
73

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/save_impl.py in __call__(self, args, *kwargs)
539 def __call__(self, args, *kwargs):
540 if not self.call_collection.tracing:
--> 541 self.call_collection.add_trace(args, *kwargs)
542 return super(LayerCall, self).__call__(args, *kwargs)
543

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/save_impl.py in wrap_with_training_arg(args, *kwargs)
482 kwargs = kwargs.copy()
483 utils.remove_training_arg(self._training_arg_index, args, kwargs)
--> 484 return call_fn(args, *kwargs)
485
486 return tf_decorator.make_decorator(

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/saving/saved_model/save_impl.py in __call__(self, args, *kwargs)
540 if not self.call_collection.tracing:
541 self.call_collection.add_trace(args, *kwargs)
--> 542 return super(LayerCall, self).__call__(args, *kwargs)
543
544 def get_concrete_function(self, args, *kwargs):

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/eager/def_function.py in __call__(self, args, *kwds)
578 xla_context.Exit()
579 else:
--> 580 result = self._call(args, *kwds)
581
582 if tracing_count == self._get_tracing_count():

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/eager/def_function.py in _call(self, args, *kwds)
616 # In this case we have not created variables on the first call. So we can
617 # run the first trace but we should fail if variables are created.
--> 618 results = self._stateful_fn(args, *kwds)
619 if self._created_variables:
620 raise ValueError("Creating variables on a non-first call to a function"

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/eager/function.py in __call__(self, args, *kwargs)
2417 """Calls a graph function specialized to the inputs."""
2418 with self._lock:
-> 2419 graph_function, args, kwargs = self._maybe_define_function(args, kwargs)
2420 return graph_function._filtered_call(args, kwargs) # pylint: disable=protected-access
2421

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/engine/base_layer.py in get_losses_for(self, inputs)
1625 losses = [l for l in self.losses if not l._unconditional_loss]
1626 inputs = nest.flatten(inputs)
-> 1627 reachable = tf_utils.get_reachable_from_inputs(inputs, losses)
1628 return [l for l in losses if l in reachable]
1629

~/anaconda3/envs/TF_2_2/lib/python3.8/site-packages/tensorflow/python/keras/utils/tf_utils.py in get_reachable_from_inputs(inputs, targets)
138 outputs = x.consumers()
139 else:
--> 140 raise TypeError('Expected Operation, Variable, or Tensor, got ' + str(x))
141
142 for y in outputs:

TypeError: Expected Operation, Variable, or Tensor, got block4

4. Expected behavior

A clear and concise description of what you expected to happen.
Differs from:
https://www.tensorflow.org/tutorials/keras/save_and_load
and related documentation

5. Additional context

Include any logs that would be helpful to diagnose the problem.

6. System information

OS Platform and Distribution (e.g., Linux Ubuntu 16.04):
Mobile device name if the issue happens on a mobile device:
TensorFlow installed from (source or binary):
TensorFlow version (use command below):
Python version:
Bazel version (if compiling from source):
GCC/Compiler version (if compiling from source):
CUDA/cuDNN version:
GPU model and memory:

print(tf.version.GIT_VERSION, tf.version.VERSION)

unknown 2.2.0
(installed via conda : in clean environment ("TF_2_2" for purpose)
conda install -c anaconda tensorflow-gpu
on Ubuntu 18.04.4 LTS
Python 3..8.3

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2017 NVIDIA Corporation
Built on Fri_Nov__3_21:07:56_CDT_2017
Cuda compilation tools, release 9.1, V9.1.85
Intel(R) Core(TM) i7-9700 CPU @ 3.00GHz
MemTotal: 65886576 kB
MemFree: 20100768 kB
MemAvailable: 56891796 kB

research support

Source

cattmi

Most helpful comment

The detection_model here is a python instance (SSDMetaArch) instead of Keras model instance, so you can't use .save or keras.save_model to save it.

pkulzc on 15 Jul 2020

👍2

All 10 comments

@cattmi

Please, fill issue template..

I am not able to open the link you shared.Request you to share colab link or complete code snippet with supporting files to reproduce the issue in our environment.It helps us in localizing the issue faster.Thanks!

ravikyram on 14 Jul 2020

I have the same problem.

Running on Colab, see notebook below.

To easily reproduce:

open the Training Tutorial Notebook
run all
create a new cell at the end with the following

detection_model.build((640, 640, 3))  
tf.keras.models.save_model(detection_model, 'savemodel')

My error message:

TypeError                                 Traceback (most recent call last)
<ipython-input-187-08c2245c2768> in <module>()
----> 1 tf.keras.models.save_model(detection_model, 'savedmodel_batch32_1000step')

78 frames
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/utils/tf_utils.py in get_reachable_from_inputs(inputs, targets)
    138       outputs = x.consumers()
    139     else:
--> 140       raise TypeError('Expected Operation, Variable, or Tensor, got ' + str(x))
    141 
    142     for y in outputs:

TypeError: Expected Operation, Variable, or Tensor, got block4

tzekid on 14 Jul 2020

I have tried in colab with TF version 2.2 and was able to reproduce the issue.Please, find the gist here.Thanks!

ravikyram on 14 Jul 2020

The detection_model here is a python instance (SSDMetaArch) instead of Keras model instance, so you can't use .save or keras.save_model to save it.

pkulzc on 15 Jul 2020

👍2

Aha..

Many Thanks for promptly resolving ..

Mike

On 15 Jul 2020, at 00:07, pkulzc notifications@github.com wrote:

The detection_model here is a python instance (SSDMetaArch) instead of Keras model instance, so you can't use .save or keras.save_model to save it.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub https://github.com/tensorflow/models/issues/8862#issuecomment-658455968, or unsubscribe https://github.com/notifications/unsubscribe-auth/AHDLGM3RB4IBXPDFW25JDW3R3TQMRANCNFSM4OZFKXIQ.

cattmi on 15 Jul 2020

In the previous tensorflow 1.x object detection api there was an option to export the model for tflite inference after removing the post processing stage. Is there a similar functionality for the new exporter? How to export a model till a specific node? I see that the saved_model when loaded using either tf.saved_model.load or keras.load haven't got the prune function, is there any other similar script to do it?

sambhusuryamohan on 15 Jul 2020

👀3

@cattmi

Please, close this thread if your issue was resolved.Thanks!

ravikyram on 15 Jul 2020

@sambhusuryamohan , we should probably start a separate thread somewhere appropriate on resolving saving and loading / export to tf lite of the python coded models as this is important topic, but I will close this thread as 'resolved': i.e. we know model.save can only work with keras implemented, not python, models..

cattmi on 15 Jul 2020