keras custom layer conversion

[arvkr]

Platform (like ubuntu 16.04/win10): Ubuntu 16.04

Python version: Python3.5

Source framework with version (like Tensorflow 1.4.1 with GPU): Keras 2.2.4

Destination framework with version (like CNTK 2.3 with GPU): IR (eventually Pytorch 1.2.0)

Pre-trained model path (webpath or webdisk path):

Running scripts: mmtoir -f keras -d ir_best_model -n best_model.json -w best_model_weights.h5

Hi,
I am trying to convert a Keras model to IR. I am getting an Unknown layer error. The layer is a custom layer implemented in Keras. My question - Can MMdnn handle custom Keras layers for conversion?

Detailed error info:

Traceback (most recent call last):
  File "/usr/local/lib/python3.5/dist-packages/mmdnn/conversion/keras/keras2_parser.py", line 69, in _load_model
    from keras.applications.mobilenet import relu6
ImportError: cannot import name 'relu6'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/usr/local/bin/mmtoir", line 11, in <module>
    sys.exit(_main())
  File "/usr/local/lib/python3.5/dist-packages/mmdnn/conversion/_script/convertToIR.py", line 192, in _main
    ret = _convert(args)
  File "/usr/local/lib/python3.5/dist-packages/mmdnn/conversion/_script/convertToIR.py", line 46, in _convert
    parser = Keras2Parser(model)
  File "/usr/local/lib/python3.5/dist-packages/mmdnn/conversion/keras/keras2_parser.py", line 126, in __init__
    model = self._load_model(model[0], model[1])
  File "/usr/local/lib/python3.5/dist-packages/mmdnn/conversion/keras/keras2_parser.py", line 78, in _load_model
    'DepthwiseConv2D': layers.DepthwiseConv2D})
  File "/usr/local/lib/python3.5/dist-packages/keras/engine/saving.py", line 492, in model_from_json
    return deserialize(config, custom_objects=custom_objects)
  File "/usr/local/lib/python3.5/dist-packages/keras/layers/__init__.py", line 55, in deserialize
    printable_module_name='layer')
  File "/usr/local/lib/python3.5/dist-packages/keras/utils/generic_utils.py", line 145, in deserialize_keras_object
    list(custom_objects.items())))
  File "/usr/local/lib/python3.5/dist-packages/keras/engine/network.py", line 1022, in from_config
    process_layer(layer_data)
  File "/usr/local/lib/python3.5/dist-packages/keras/engine/network.py", line 1008, in process_layer
    custom_objects=custom_objects)
  File "/usr/local/lib/python3.5/dist-packages/keras/layers/__init__.py", line 55, in deserialize
    printable_module_name='layer')
  File "/usr/local/lib/python3.5/dist-packages/keras/utils/generic_utils.py", line 138, in deserialize_keras_object
    ': ' + class_name)
ValueError: Unknown layer: BilinearUpSampling2D

Thanks in advance!

[rainLiuplus]

Hi @arvkr, Now MMdnn can not handle the custom keras layer automatically. But if you still want to convert your model via MMdnn, you could:

1. Modify the MMdnn code to load your custom model.
   Please refer to this to change the load model code.

2. Add the custom layer.
   Please refer to contribution guideline.

Thanks!

[arvkr]

Thank you for this info! I will try this method and hopefully it works!

[arvkr]

I converted the custom layer to an inbuilt Keras layer and converted the model successfully to pytorch. I did not have to resort to modifying the Keras parser after all.