Handling multiple inputs in keras

[sRassmann]

Platform (like ubuntu 16.04/win10): Ubuntu 18.04 (Google Colab)

Python version: 3.7.10

Source framework with version (like Tensorflow 1.4.1 with GPU): Keras 2.5 with Tensorflow 2.0 GPU Backend

Pre-trained model path (webpath or webdisk path): relevant model config JSON

Destination framework with version (like CNTK 2.3 with GPU): PyTorch 1.6.0 GPU

I would like to convert an existing (trained) model from keras/tf to pytorch.
However, the model uses two inputs (an image and an additional boolean value) and, hence is currently implemented as keras.engine.functional.Functional. Apparently, MMdnn can not handle it:

$ mmconvert -sf keras -iw output/model.h5 -df pytorch -om output/model.pth
$ # and also with
$ mmtoir -f keras -d output/dbam -n data/models/tf/dbam.json
> Traceback (most recent call last):
  File "/usr/local/bin/mmtoir", line 8, in <module>
    sys.exit(_main())
  File "/usr/local/lib/python3.7/dist-packages/mmdnn/conversion/_script/convertToIR.py", line 197, in _main
    ret = _convert(args)
  File "/usr/local/lib/python3.7/dist-packages/mmdnn/conversion/_script/convertToIR.py", line 46, in _convert
    parser = Keras2Parser(model)
  File "/usr/local/lib/python3.7/dist-packages/mmdnn/conversion/keras/keras2_parser.py", line 135, in __init__
    self.keras_graph = Keras2Graph(model)
  File "/usr/local/lib/python3.7/dist-packages/mmdnn/conversion/keras/keras2_graph.py", line 37, in __init__
    raise TypeError("Keras layer of type %s is not supported." % type(model))
TypeError: Keras layer of type <class 'keras.engine.functional.Functional'> is not supported.

Is there any alternative way for trying the conversion that I missed?
Else is there any workaround to this?

Here is the code to generate the keras model:

img = Input(shape=image_shape)
gender = Input(shape=(1,))
cnn_vec = InceptionV3(input_shape=image_shape, include_top=False, weights=None)(img)
cnn_vec = GlobalAveragePooling2D()(cnn_vec)
cnn_vec = Dropout(0.2)(cnn_vec)
gender_vec = Dense(32, activation="relu")(gender)
features = Concatenate(axis=-1)([cnn_vec, gender_vec])
dense_layer = Dense(1024, activation="relu")(features)
dense_layer = Dropout(0.2)(dense_layer)
dense_layer = Dense(1024, activation="relu")(dense_layer)
dense_layer = Dropout(0.2)(dense_layer)
dense_layer = Dense(512, activation="relu")(dense_layer)
dense_layer = Dropout(0.2)(dense_layer)
dense_layer = Dense(512, activation="relu")(dense_layer)
dense_layer = Dropout(0.2)(dense_layer)
output_layer = Dense(1, activation="linear")(dense_layer)
model = Model(inputs=[img, gender], outputs=output_layer)

adam = optimizers.Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=0.0)

model.compile(optimizer=adam, loss="mse", metrics=metrics)

Thanks for your help.