[OpenVINO-EP] Using Core::ReadNetwork() method for creating a CNNNetwork

[MaajidKhan]

Description:
Using Core::ReadNetwork() method for reading and creating a CNNNetwork for OpenVINO versions >=2020.4

Motivation and Context
Since OpenVINO™ 2020.4 version, Inference Engine enables reading ONNX models via the Inference Engine Core API and there is no need to use directly the low-level ONNX* Importer API anymore. To read ONNX* models, it's recommended to use the Core::ReadNetwork() method that provide a uniform way to read models from ONNX format.

More Info here:
ONNX* Importer API is deprecated.
https://docs.openvinotoolkit.org/latest/openvino_docs_IE_DG_OnnxImporterTutorial.html

Info on ReadNetwork() method from InferenceEngine::Core Class:
https://docs.openvinotoolkit.org/latest/classInferenceEngine_1_1Core.html#a251861e52a979d6e61848babae3673ef

[HectorSVC]

/azp run Linux CPU CI Pipeline,Linux CPU x64 NoContribops CI Pipeline,Linux GPU CI Pipeline,Linux GPU TensorRT CI Pipeline,MacOS CI Pipeline,MacOS NoContribops CI Pipeline,Windows CPU CI Pipeline,Windows GPU CI Pipeline,Windows GPU TensorRT CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 9 pipeline(s).

[HectorSVC]

/azp run orttraining-linux-ci-pipeline,orttraining-mac-ci-pipeline,orttraining-linux-gpu-ci-pipeline,centos7_cpu,Linux OpenVINO CI Pipeline,/azp run orttraining-linux-ci-pipeline,orttraining-mac-ci-pipeline,orttraining-linux-gpu-ci-pipeline,centos7_cpu,Linux OpenVINO CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 5 pipeline(s).

[HectorSVC]

/azp run Linux CPU Minimal Build E2E CI Pipeline, orttraining-distributed

[azure-pipelines]

Azure Pipelines successfully started running 2 pipeline(s).

[HectorSVC]

@MaajidKhan, please take a look at the test failure
Test #1: onnxruntime_test_all ...................***Exception: SegFault924.77 sec

[MaajidKhan]

Hector, I am looking into this issue. I will let you know once I fix that.

[HectorSVC]

/azp run Linux OpenVINO CI Pipeline

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[MaajidKhan]

@HectorSVC
so, when we are trying to use ReadNetwork() method to create the CNNNetwork instead of the low-level ONNX* Importer API and try running any c++ application, we get a seg fault issue. We suspect this issue is because of protobuf library mismatch or something. I think the problem can even be caused by different runtime versions. I mean when we use another compiler to build OpenVINO and you get SEGFAULT due to ABI problems.

This issue is only when we are building onnxruntime with OpenVINO Binary Release package. However, we are able to successfully run c++ applications when onnxruntime is built with Openvino from source ( master (or) releases/2021/2 branch).

so, we think, When we are building openvino from source, the protobuf gets installed on the fly and hence we are not facing any issues and this is the same with onnxruntime.

In onnxruntime, the segfault happens at this line of code where it tries to shutdown the protobuf libraray.
https://github.com/microsoft/onnxruntime/blob/master/onnxruntime/test/perftest/main.cc#L73

we are talking to OpenVINO team to check how this issue could be resolved. so closing this PR for now, will create a new one, once we get this thing resolved.