Converting pytorch model with multiple inputs to Openvino

[EBBIAV]

Note: For some reason my last question immediately got marked as spam. That's why I repost this question.

I already had a look at this very similar issue, unfortunately the solution did not work out for me. My model is also developed in pytorch and has two inputs. When I convert from ONNX to Openvino I get the following error:

[ ERROR ] Exception occurred during running replacer "REPLACEMENT_ID" (<class 'extensions.front.user_data_repack.UserDataRepack'>): No node with name 0

My setup is like this:  

torch.onnx.export(
    model,
    (dummy_input_faces, dummy_input_face_marks),
    f'{OUTPUT_DIR.as_posix()}/{model.__class__.__name__}.onnx',
    input_names=["0", "1"],
    opset_version=9,
    do_constant_folding=True
)

# Get the path to the Model Optimizer script
mo_path = str(Path(mo_onnx.__file__))
# Construct the command for Model Optimizer
print(OUTPUT_DIR / model.__class__.__name__)
mo_command = f""""{sys.executable}"
                 "{mo_path}"
                 --input_model "{OUTPUT_DIR / model.__class__.__name__}.onnx"
                 --input "0,1"
                 --input_shape "(1,256,256,3), (1,68,2)"
                 --data_type FP16
                 --output_dir "{OUTPUT_DIR}"
                 """

This is what the above command returns:

Model Optimizer arguments:
Common parameters:
- Path to the Input Model: C:\Users\I008619\Projects\depth_ai\conversion\outputs\PreprocessAnon.onnx
- Path for generated IR: C:\Users\I008619\Projects\depth_ai\conversion\outputs
- IR output name: PreprocessAnon
- Log level: ERROR
- Batch: Not specified, inherited from the model
- Input layers: 0,1
- Output layers: Not specified, inherited from the model
- Input shapes: (1,256,256,3), (1,68,2)
- Mean values: Not specified
- Scale values: Not specified
- Scale factor: Not specified
- Precision of IR: FP16
- Enable fusing: True
- Enable grouped convolutions fusing: True
- Move mean values to preprocess section: None
- Reverse input channels: False
ONNX specific parameters:
- Inference Engine found in: C:\Users\I008619\.virtualenvs\openvino_env\lib\site-packages\openvino
Inference Engine version: 2021.4.0-3839-cd81789d294-releases/2021/4
Model Optimizer version: 2021.4.0-3839-cd81789d294-releases/2021/4
[ ERROR ] Exception occurred during running replacer "REPLACEMENT_ID" (<class 'extensions.front.user_data_repack.UserDataRepack'>): No node with name 0

Can anyone help me with this? I could also share the ONNX model if needed.

[Zulkifli_Intel]

Hello Enrico Boos,

Thank you for reaching out to us.

Please share with us your model. In the meantime, please try to convert the model without using any parameters and see if the model can be converted to IR format. Also, check this Converting a PyTorch* Model documentation for additional information.

Sincerely,

Zulkifli

[EBBIAV]

I had to post the question multiple times, because it was always marked as spam. Please delete this question

[Zulkifli_Intel]

Hello Enrico,

We will close this thread. If you need any additional information from Intel, please submit a new question and we'll be happy to help you.

Sincerely,

Zulkifli