onnx
convert onnx model to coreml model error?
when I run the demo utorials/examples/CoreML/ONNXLive/,step by step,I convert onnx model to coreml model,
python onnx_to_coreml.py ./saved_models/candy.onnx ./saved_models/candy.mlmodel
I shows the error message like
51/64: Converting Node Type InstanceNormalization 52/64: Converting Node Type Add 53/64: Converting Node Type Upsample Traceback (most recent call last): File "onnx_to_coreml.py", line 22, in <module> coreml_model = convert(model_proto, image_input_names=['0'], image_output_names=['186']) File "/Users/lixianjie/anaconda2/envs/cpp/lib/python3.6/site-packages/onnx_coreml/converter.py", line 458, in convert _convert_node(builder, node, graph, err) File "/Users/lixianjie/anaconda2/envs/cpp/lib/python3.6/site-packages/onnx_coreml/_operators.py", line 1755, in _convert_node return converter_fn(builder, node, graph, err) File "/Users/lixianjie/anaconda2/envs/cpp/lib/python3.6/site-packages/onnx_coreml/_operators.py", line 1440, in _convert_upsample scales = node.attrs["scales"] KeyError: 'scales'
can anybody tell me what's wrong with my code?
just same error with yours. by the way, I use the Netron to see the model structure and the final ID is 128, not the 186 which the example mentioned.
below is my error log:WARNING:root:TensorFlow version 1.8.0 detected. Last version known to be fully compatible is 1.5.0 .
1/64: Converting Node Type Pad
2/64: Converting Node Type Conv
3/64: Converting Node Type InstanceNormalization
4/64: Converting Node Type Relu
5/64: Converting Node Type Pad
6/64: Converting Node Type Conv
7/64: Converting Node Type InstanceNormalization
8/64: Converting Node Type Relu
9/64: Converting Node Type Pad
10/64: Converting Node Type Conv
11/64: Converting Node Type InstanceNormalization
12/64: Converting Node Type Relu
13/64: Converting Node Type Pad
14/64: Converting Node Type Conv
15/64: Converting Node Type InstanceNormalization
16/64: Converting Node Type Relu
17/64: Converting Node Type Pad
18/64: Converting Node Type Conv
19/64: Converting Node Type InstanceNormalization
20/64: Converting Node Type Add
21/64: Converting Node Type Pad
22/64: Converting Node Type Conv
23/64: Converting Node Type InstanceNormalization
24/64: Converting Node Type Relu
25/64: Converting Node Type Pad
26/64: Converting Node Type Conv
27/64: Converting Node Type InstanceNormalization
28/64: Converting Node Type Add
29/64: Converting Node Type Pad
30/64: Converting Node Type Conv
31/64: Converting Node Type InstanceNormalization
32/64: Converting Node Type Relu
33/64: Converting Node Type Pad
34/64: Converting Node Type Conv
35/64: Converting Node Type InstanceNormalization
36/64: Converting Node Type Add
37/64: Converting Node Type Pad
38/64: Converting Node Type Conv
39/64: Converting Node Type InstanceNormalization
40/64: Converting Node Type Relu
41/64: Converting Node Type Pad
42/64: Converting Node Type Conv
43/64: Converting Node Type InstanceNormalization
44/64: Converting Node Type Add
45/64: Converting Node Type Pad
46/64: Converting Node Type Conv
47/64: Converting Node Type InstanceNormalization
48/64: Converting Node Type Relu
49/64: Converting Node Type Pad
50/64: Converting Node Type Conv
51/64: Converting Node Type InstanceNormalization
52/64: Converting Node Type Add
53/64: Converting Node Type Upsample
Traceback (most recent call last):
File "onnx_to_coreml.py", line 35, in
problem has been solved by this link: https://github.com/onnx/onnx-coreml/issues/365#issuecomment-459389469