Having issue in converting autokeras functional model to onnx?

[hanzigs]

Hi, I am trying to convert autokeras model to onnx, i get error

from autokeras import StructuredDataClassifier
model = StructuredDataClassifier(max_trials=100)
model.fit(x=X_train, y=y_train, validation_data=(X_valid, y_valid), epochs=1000, verbose=1)
autoKeras_model = model.export_model()
autoKeras_model.summary()

Model: "functional_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         [(None, 18)]              0         
_________________________________________________________________
multi_category_encoding (Mul (None, 18)                0         
_________________________________________________________________
dense (Dense)                (None, 32)                608       
_________________________________________________________________
re_lu (ReLU)                 (None, 32)                0         
_________________________________________________________________
dense_1 (Dense)              (None, 32)                1056      
_________________________________________________________________
re_lu_1 (ReLU)               (None, 32)                0         
_________________________________________________________________
dense_2 (Dense)              (None, 1)                 33        
_________________________________________________________________
classification_head_2 (Activ (None, 1)                 0         
=================================================================
Total params: 1,697
Trainable params: 1,697
Non-trainable params: 0
_________________________________________________________________

converting to onnx

import onnxruntime
import keras2onnx
onnx_model = keras2onnx.convert_keras(autoKeras_model, "autokeras", debug_mode=1)

I get error

tf executing eager_mode: True
INFO:keras2onnx:tf executing eager_mode: True
tf.keras model eager_mode: False
INFO:keras2onnx:tf.keras model eager_mode: False
Model: "functional_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         [(None, 18)]              0         
_________________________________________________________________
multi_category_encoding (Mul (None, 18)                0         
_________________________________________________________________
dense (Dense)                (None, 32)                608       
_________________________________________________________________
re_lu (ReLU)                 (None, 32)                0         
_________________________________________________________________
dense_1 (Dense)              (None, 32)                1056      
_________________________________________________________________
re_lu_1 (ReLU)               (None, 32)                0         
_________________________________________________________________
dense_2 (Dense)              (None, 1)                 33        
_________________________________________________________________
classification_head_2 (Activ (None, 1)                 0         
=================================================================
Total params: 1,697
Trainable params: 1,697
Non-trainable params: 0
_________________________________________________________________
None
Traceback (most recent call last):

  File "<ipython-input-76-0567e6de6858>", line 1, in <module>
    onnx_model = keras2onnx.convert_keras(ExportedautoKeras_model, model_name, debug_mode=1)

  File "C:\Users\Pe\Anaconda3\lib\site-packages\keras2onnx\main.py", line 62, in convert_keras
    tf_graph = build_layer_output_from_model(model, output_dict, input_names, output_names)

  File "C:\Users\Pe\Anaconda3\lib\site-packages\keras2onnx\_parser_tf.py", line 302, in build_layer_output_from_model
    return extract_outputs_from_subclassing_model(model, output_dict, input_names, output_names)

  File "C:\Users\Pe\Anaconda3\lib\site-packages\keras2onnx\_parser_tf.py", line 264, in extract_outputs_from_subclassing_model
    concrete_func, lower_control_flow=True)

  File "C:\Users\Pe\Anaconda3\lib\site-packages\keras2onnx\_graph_cvt.py", line 437, in convert_variables_to_constants_v2
    tensor_data = _get_tensor_data(func)

  File "C:\Users\Pe\Anaconda3\lib\site-packages\keras2onnx\_graph_cvt.py", line 209, in _get_tensor_data
    data = val_tensor.numpy()

  File "C:\Users\Pe\Anaconda3\lib\site-packages\tensorflow\python\framework\ops.py", line 1063, in numpy
    maybe_arr = self._numpy()  # pylint: disable=protected-access

  File "C:\Users\Pe\Anaconda3\lib\site-packages\tensorflow\python\framework\ops.py", line 1031, in _numpy
    six.raise_from(core._status_to_exception(e.code, e.message), None)  # pylint: disable=protected-access

  File "<string>", line 3, in raise_from

InvalidArgumentError: Cannot convert a Tensor of dtype resource to a NumPy array.

May i have some help Thanks

[hanzigs]

Hi, Can i have some help on the above please, Using tensorflow == 2.3.0 keras == 2.4.3 Thanks

[hanzigs]

Hi, Can i have some help on the above please,

import tensorflow as tf
converter = tf.lite.TFLiteConverter.from_keras_model(autoKeras_model)
tflite_model = converter.convert()

tried this, same error