Error converting model from keras using tensorflow-gpu backend

$pip list onnx 1.4.1 onnxmltools 1.3.2 tensorflow 1.13.1 tensorflow-gpu 1.13.1

$conda list keras 2.2.4

import onnxmltools
from keras.layers import Input, Dense, Add
from keras.models import Model

# N: batch size, C: sub-model input dimension, D: final model's input dimension
N, C, D = 2, 3, 3

# Define a sub-model, it will become a part of our final model
sub_input1 = Input(shape=(C,))
sub_mapped1 = Dense(D)(sub_input1)
sub_model1 = Model(inputs=sub_input1, outputs=sub_mapped1)

# Define another sub-model, it will become a part of our final model
sub_input2 = Input(shape=(C,))
sub_mapped2 = Dense(D)(sub_input2)
sub_model2 = Model(inputs=sub_input2, outputs=sub_mapped2)

# Define a model built upon the previous two sub-models
input1 = Input(shape=(D,))
input2 = Input(shape=(D,))
mapped1_2 = sub_model1(input1)
mapped2_2 = sub_model2(input2)
sub_sum = Add()([mapped1_2, mapped2_2])
keras_model = Model(inputs=[input1, input2], output=sub_sum)

# Convert it! The target_opset parameter is optional.
onnx_model = onnxmltools.convert_keras(keras_model, target_opset=7) 

Error:

C:\Users\intel\Anaconda3\envs\denoiser\lib\site-packages\ipykernel_launcher.py:24: UserWarning: Update your `Model` call to the Keras 2 API: `Model(inputs=[<tf.Tenso..., outputs=Tensor("ad...)`
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-7-13c32f7fd452> in <module>
     25 
     26 # Convert it! The target_opset parameter is optional.
---> 27 onnx_model = onnxmltools.convert_keras(keras_model, target_opset=7)

~\AppData\Roaming\Python\Python36\site-packages\onnxmltools\convert\main.py in convert_keras(model, name, initial_types, doc_string, target_opset, targeted_onnx, channel_first_inputs, custom_conversion_functions, custom_shape_calculators, default_batch_size)
     30 
     31     from keras2onnx import convert_keras as convert
---> 32     return convert(model, name, doc_string, target_opset, channel_first_inputs)
     33 
     34 

~\AppData\Roaming\Python\Python36\site-packages\keras2onnx\main.py in convert_keras(model, name, doc_string, target_opset, channel_first_inputs, debug_mode, custom_op_conversions)
     92     tf_graph_def = tf.graph_util.convert_variables_to_constants(sess, sess.graph_def, output_node_names=out_node)
     93     return _convert_tf(name, tf_graph_def, op_dict, output_names, target_opset, doc_string, channel_first_inputs,
---> 94                        debug_mode, custom_op_conversions)
     95 
     96 

~\AppData\Roaming\Python\Python36\site-packages\keras2onnx\main.py in _convert_tf(name, tf_graph_def, keras_op_table, output_names, target_opset, doc_string, channel_first_inputs, debug_mode, custom_op_conversions)
     64         topology.compile()
     65 
---> 66         return convert_topology(topology, name, doc_string, target_opset, channel_first_inputs)
     67 
     68 

~\AppData\Roaming\Python\Python36\site-packages\keras2onnx\topology.py in convert_topology(topology, model_name, doc_string, target_opset, channel_first_inputs)
    232         scope = next(scope for scope in topology.scopes if scope.name == operator.scope)
    233         keras2onnx_logger().debug("Converting the operator (%s): %s" % (operator.full_name, operator.type))
--> 234         get_converter(operator.type)(scope, operator, container)
    235 
    236     # When calling ModelComponentContainer's add_initializer(...), nothing is added into the input list. However, in

~\AppData\Roaming\Python\Python36\site-packages\keras2onnx\wrapper.py in tfnode_convert(varset, operator, container)
     51     # create input_tensor_values, initializers
     52     # if initilizer is not used as input by any node, then it will be ignored
---> 53     initializers = [i for i in list(g.initializers.values()) if i.name in all_inputs]
     54     for init_tensor_ in initializers:
     55         init_tensor_.name = varset.get_local_variable_or_declare_one(init_tensor_.name).full_name.encode('utf-8')

AttributeError: 'Graph' object has no attribute 'initializers'

[jiafatom]

@arielbernal @cyroxx @cwx123147 We just update the latest keras2onnx, please re install. Thanks. Below is my suggested workaround which you can ignore.

---

Please install keras2onnx through the GitHub one here https://github.com/onnx/keras-onnx.git Then it should work.

The reason is that keras2onnx uses an API in tensorflow-onnx, and their API made a breaking change, which causes this AttributeError. We fix it in keras2onnx, will publish a new pypi soon. So please install keras2onnx from GitHub as mentioned above.

[arielbernal]

@jiafatom @cyroxx Awesome. confirmed working!