Community
cancel
Showing results for 
Search instead for 
Did you mean: 
Highlighted
Novice
294 Views

"RuntimeError: Unexpected CNNNetwork format" with Neural Compute Stick 2 and OpenVINO

Jump to solution

Inference works well on CPU for my custom model, but I get this error when running it on a raspberry pi with a Neural Compute Stick. Snippet from log:

File "/media/pi/0403-0201/InferenceST/src/stylizer.py", line 17, in set_style
self.inferenceEngine = InferenceEngine(path_to_model)
File "/media/pi/0403-0201/InferenceST/src/inference_engine.py", line 13, in __init__
self.executable_net = self.get_executable_network(path_to_model)
File "/media/pi/0403-0201/InferenceST/src/inference_engine.py", line 20, in get_executable_network
exec_net = self.ie.load_network(network=self.net, device_name=self.device)
File "ie_api.pyx", line 178, in openvino.inference_engine.ie_api.IECore.load_network
File "ie_api.pyx", line 187, in openvino.inference_engine.ie_api.IECore.load_network
RuntimeError: Unexpected CNNNetwork format: it was converted to deprecated format prior plugin's call

 

The error seems to be related to the "input" layer in the model which is supported by the CPU but not the MYRIAD NCS. 

Things I have tried:

    -  using --data_type FP16 with the model optimizer

    - Removing the "input" layer through the model optimizer using the "--input" argument.

    - using the HETERO Plugin but this doesn't work on the Raspberry pi

This seems like a solvable problem but I'm not sure what to try. Should I implement a custom layer? Is the input layer even the cause of the problem? Any help would be appreciated.

Thanks in advance,

gilgo

Labels (3)
0 Kudos

Accepted Solutions
Highlighted
Employee
272 Views

Hi gilgo,

Thanks for reaching out. Could you please answer the following:

  • Which OpenVINO™ toolkit version are you using on the Rpi?
  • Could you please tell us which sample are you using and share the model and the command used to replicate your issue?

Please try to convert the model using the latest version of OpenVINO™ toolkit and use the latest version as well to run the sample on the Rpi. If you are using different versions, there might be an incompatibility issue.

Regards,

David C.


View solution in original post

0 Kudos
3 Replies
Highlighted
Employee
273 Views

Hi gilgo,

Thanks for reaching out. Could you please answer the following:

  • Which OpenVINO™ toolkit version are you using on the Rpi?
  • Could you please tell us which sample are you using and share the model and the command used to replicate your issue?

Please try to convert the model using the latest version of OpenVINO™ toolkit and use the latest version as well to run the sample on the Rpi. If you are using different versions, there might be an incompatibility issue.

Regards,

David C.


View solution in original post

0 Kudos
Highlighted
Novice
263 Views

Hi David,

I was actually using 2020.3 on the pi and 2020.4 for the model optimizer. Changing to 2020.4 on the pi solved my issue.

Thanks a lot!

gilgo

0 Kudos
Highlighted
Employee
258 Views

Hi gilgo,

Happy to hear you solved it! This thread will no longer be monitored, so in case you have additional questions feel free to post a new discussion and we will help you there.

Best regards,

David C.


0 Kudos