Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Highlighted
New Contributor I
95 Views

ONNX conversion - ops missing

Hello

I am trying to convert a CNN model from ONNX to OpenVINO.

The following 2 ops are missing:

Instructions/layers that do not have attribute extractors:
[ WARNING ]      Sub (2)
[ WARNING ]          Sub_78
[ WARNING ]          Sub_140
[ WARNING ]      Min (3)
[ WARNING ]          Min_87
[ WARNING ]          Min_88
[ WARNING ]          Min_89

Min and Sub. Can you please provide some guidance as to how/where to add them?

Thank you

0 Kudos
3 Replies
Highlighted
Moderator
79 Views

Hi Ra Tr,

 

Sub should be a supported ONNX layer while Min is not.

You can check ONNX supported layers to see a list of supported layers.

 

There is a section for implement custom layer using the Custom Layers Guide.

However, it seems that ONNX conversion could not support custom layers.

 

Therefore, I would recommend you to tweak your model to not use the Min layer if possible.

 

Additionally, here is a list of validated models and topologies that are compatible.

 

Regards,

Rizal


0 Kudos
Highlighted
Moderator
64 Views

Hi Ra,


Intel will no longer monitor this thread since we have provided a solution. If you need any additional information from Intel, please submit a new question.


Regards,

Rizal


0 Kudos
Highlighted
New Contributor I
59 Views

Is this a naming problem?
I see Minimum is supported here: https://docs.openvinotoolkit.org/latest/openvino_docs_ops_arithmetic_Minimum_1.html

0 Kudos