[Slice of Caffe parser support] Error when converting DPN68 from caffe to IR

[seanchung2]

Hi,

This is my model: link

The error message showed up when I tried to convert my model from caffe to IR:

Traceback (most recent call last):
  File "/usr/lib/python2.7/runpy.py", line 174, in _run_module_as_main
    "__main__", fname, loader, pkg_name)
  File "/usr/lib/python2.7/runpy.py", line 72, in _run_code
    exec code in run_globals
  File "/usr/local/lib/python2.7/dist-packages/mmdnn/conversion/_script/convertToIR.py", line 159, in <module>
    _main()
  File "/usr/local/lib/python2.7/dist-packages/mmdnn/conversion/_script/convertToIR.py", line 154, in _main
    ret = _convert(args)
  File "/usr/local/lib/python2.7/dist-packages/mmdnn/conversion/_script/convertToIR.py", line 9, in _convert
    transformer = CaffeTransformer(args.network, args.weights, "tensorflow", args.inputShape, phase = args.caffePhase)
  File "/usr/local/lib/python2.7/dist-packages/mmdnn/conversion/caffe/transformer.py", line 308, in __init__
    graph = GraphBuilder(def_path, self.input_shape, self.is_train_proto, phase).build()
  File "/usr/local/lib/python2.7/dist-packages/mmdnn/conversion/caffe/graph.py", line 444, in build
    graph.compute_output_shapes(self.model)
  File "/usr/local/lib/python2.7/dist-packages/mmdnn/conversion/caffe/graph.py", line 265, in compute_output_shapes
    node.output_shape = TensorShape(*NodeKind.compute_output_shape(node))
  File "/usr/local/lib/python2.7/dist-packages/mmdnn/conversion/caffe/graph.py", line 127, in compute_output_shape
    raise ConversionError('Output shape computation not implemented for type: %s' % node.kind)
mmdnn.conversion.caffe.errors.ConversionError: Output shape computation not implemented for type: Slice

Can you check this out?

Thank you

[kitstar]

Slice layer of Caffe parser is not implemented. We will inform you asap. Thanks.

[gafr]

any update on this one?

[Alan-Turing-Ko]

Any update?

[l8rpeace]

Any update on this?

[ljie-PI]

any update on this :)

[AllenMao]

any update on this :)

[BasicCoder]

I have an immature solution in caffe to tensorflow: in mmdnn/conversion/caffe/shape.py add code:

def shape_slice(node):    
    print(node.parameters)    
    axis = node.parameters.axis    
    slices = len(node.parameters.slice_point) if len(node.parameters.slice_point) > 2 else 2    
    input_shape = node.get_only_parent()[0].output_shape    
    if axis == 1:    
        return TensorShape(input_shape.batch_size, input_shape.channels // slices, input_shape.height, input_shape.width)    
    elif axis == 2:    
        return TensorShape(input_shape.batch_size, input_shape.channels, input_shape.height // slices, input_shape.width)    
    elif axis == 3:    
        return TensorShape(input_shape.batch_size, input_shape.channels, input_shape.height, input_shape.width // slices)    
    return TensorShape(input_shape.batch_size, input_shape.channels // slices, input_shape.height, input_shape.width)    

in mmdnn/conversion/caffe/graph.py modify code:

...    
'Split': shape_split,    
'Slice': shape_slice,    
'TanH': shape_identity,    
...    

in mmdnn/conversion/caffe/mapper.py add code in class NodeMapper:

@classmethod    
 def map_slice(cls, node):    
       slices = len(node.parameters.slice_point) if len(node.parameters.slice_point) > 2 else 2    
       kwargs = {'axis': (2, 3, 1, 0)[node.parameters.axis], 'num_or_size_splits': slices}    
       cls._convert_output_shape(kwargs, node)    
       return Node.create('Split', **kwargs)    

so you can run command:

mmtoir -f caffe -n ./models/model.prototxt -w ./models/model.caffemodel -o caffe_model_IR    

mmtocode -f tensorflow --IRModelPath caffe_model_IR.pb --IRWeightPath caffe_model_IR.npy --dstModelPath tf_model_IR.py

It generates tensorflow model code: tf_model_IR.py Now you can modify tf_model_IR.py source file to correspond to caffe model

Note: This can only divide input tensor evenly.