OpenVINO: MYRIAD RuntimeError

[Han__Tianyi]

I am running into some trouble when trying to use Inference Engine on the NCS2 through the python API (python 3.7.3, Host Windows 10, ONNX PyTorch)

I also used the 2020.1 version of openvino. The model is AlexNet in torchvision.

model_ft = torchvision.models.alexnet(pretrained=True) 

I don't see this error with this same model for CPU and GPU. The error is in the code of creating exe_net

exec_net = plugin.load(network=net) 

Traceback (most recent call last):
  File "C:/Users/Hans/Desktop/保研/情绪识别/demo/audio/alexnet_vote/test2.py", line 204, in <module>
    main(testdata, test_mapping)
  File "C:/Users/Hans/Desktop/保研/情绪识别/demo/audio/alexnet_vote/test2.py", line 138, in main
    exec_net = plugin.load(network=net)
  File "ie_api.pyx", line 1205, in openvino.inference_engine.ie_api.IEPlugin.load
  File "ie_api.pyx", line 1215, in openvino.inference_engine.ie_api.IEPlugin.load
RuntimeError: Check (env.config.ignoreUnknownLayers) failed: Failed to compile layer 38 : AssertionFailed: input->desc().totalDimSize() == output->desc().totalDimSize()
E: [global] [         0] [] XLinkResetRemote:243    can't wait dispatcherClosedSem

Here is the xml: 

<layer id="31" name="38" type="Reshape" version="opset1">  
    <data special_zero="True"/>  
    <input>  
        <port id="0">  
            <dim>10</dim>  
            <dim>256</dim>  
            <dim>6</dim>  
            <dim>6</dim>  
        </port>  
        <port id="1">  
            <dim>2</dim>  
        </port>  
    </input>  
    <output>  
        <port id="2" precision="FP16">  
            <dim>10</dim>  
            <dim>9216</dim>  
        </port>  
    </output>  
</layer>

[JAIVIN_J_Intel]

Hi Tianyi,

Have you converted your onnx model to FP16 IR model via OpenVINO toolkit 2020 Model Optimizer?

If not, please try to convert using the latest version and let us know the results.

If the issue still persists, can you share the model to check from our end? If required I can send a PM to share the model privately.

Regards,

Jaivin

[Han__Tianyi]

Jaivin James (Intel) wrote:

Hi Tianyi,

Have you converted your onnx model to FP16 IR model via OpenVINO toolkit 2020 Model Optimizer?

If not, please try to convert using the latest version and let us know the results.

If the issue still persists, can you share the model to check from our end? If required I can send a PM to share the model privately.

Regards,

Jaivin

hi Jaivin,

Yes, I converted your onnx model to FP16 already, and the issue still persists. How can I send the model to you.

thanks

[JAIVIN_J_Intel]

Hi Tianyi,

Have you used any previous version of OpenVINO to convert the model to IR?

I have send you a Private Message. Please share your model via the same.

Regards,

Jaivin

[Han__Tianyi]

hi Jaivin,

did you receive my private message

Regards,

Tianyi

[JAIVIN_J_Intel]

Hi Tianyi,

Thank you for sharing the model.

The issue has been reported to the Engineering team. We are working on this issue and will report back once we get an update.

Regards,

Jaivin

[Han__Tianyi]

Hi Jaivin,

Thank you for your help.

Regards,

Tianyi

[Morsali__Reza]

Hi Tianyi and Jaivin, I am experiencing the same problem. Could you let me know if you find the solution?

[Han__Tianyi]

Hi Reza,

I haven't found a solution yet, maybe we need to wait for the update.

Regards,

Tianyi

[JAIVIN_J_Intel]

The Engineering team is working on this and related issues. Thank you for reporting the concerns that helps to make OpenVINO a better Intel software product.

[flyer]

Jaivin James (Intel) wrote:

I'm facing the same problem, still no update or solution? Thanks

E: [global] [         0] [] XLinkResetRemote:249        can't wait dispatcherClosedSem

The Engineering team is working on this and related issues. Thank you for reporting the concerns that helps to make OpenVINO a better Intel software product.

[KubaB]

I have the save issue with torchvision.models.resnet18. Any update?

I use:
Python 3.7.6, Host Windows 10, ONNX PyTorch, openvino 2020.3.194