[Bug]: py compile ir model success, c++ compile ir model failed!

[hhxdestiny]

OpenVINO Version

2024.2

Operating System

Windows System

Device used for inference

CPU

Framework

None

Model used

PP-OCRv4

Issue description

I converted the PaddleOcrV4 model to IR format using OVC in Python openvino2024.2, and then I loaded the model using Python openvino2024.2 normally. However, I encountered an error while loading it using C++openvino2024.2 (vcpkg install openvino).

C++ code:

#include <iostream>
#include <openvino/openvino.hpp>

int main(int argc, char* argv[]) {
  ov::Core core;
  auto versions = core.get_versions("CPU");
  for (const auto& version : versions) {
    std::cout << version.first << ": " << version.second << '\n';
  }
  auto model = core.read_model("model.xml");
  try {
    auto compiled_model = core.compile_model(model, "AUTO");
  } catch (const std::exception& e) {
    std::cerr << e.what() << '\n';
  }
}

Python code:

import openvino as ov

core = ov.Core()
versions = core.get_versions("CPU")
print(versions)
model = core.read_model(model="model.xml")
print(model)
compiled_model = core.compile_model(model=model, device_name="AUTO")
print(compiled_model)

Step-by-step reproduction

No response

Relevant log output

C++ output:

CPU: openvino_intel_cpu_plugin
    Version : 2024.2.0
    Build   : 2024.2.0-000--

Exception from src\inference\src\cpp\core.cpp:107:
Exception from src\inference\src\dev\plugin.cpp:54:
Exception from src\plugins\auto\src\auto_schedule.cpp:431:
[AUTO] compile model failed, CPU:Exception from src\inference\src\dev\plugin.cpp:54:
bad combination
;

Python output:

{'CPU': <Version: 2024.2.0-15519-5c0f38f83f6-releases/2024/2 openvino_intel_cpu_plugin>}
<Model: 'Model0'
inputs[
<ConstOutput: names[x] shape[?,3,?,?] type: f32>
]
outputs[
<ConstOutput: names[sigmoid_0.tmp_0] shape[?,1,32..,32..] type: f32>
]>
<CompiledModel:
inputs[
<ConstOutput: names[x] shape[?,3,?,?] type: f32>
]
outputs[
<ConstOutput: names[sigmoid_0.tmp_0] shape[?,1,32..,32..] type: f32>
]>

Issue submission checklist

• [X] I'm reporting an issue. It's not a question.
• [X] I checked the problem with the documentation, FAQ, open issues, Stack Overflow, etc., and have not found a solution.
• [X] There is reproducer code and related data files such as images, videos, models, etc.

[hhxdestiny]

Other ir models also have this issue

[andrei-kochin]

@hhxdestiny thank you for reaching the OpenVINO!

Could you please check if the issue is present when you explicitly specify the device like below?

auto compiled_model = core.compile_model(model, "CPU"); 

[hhxdestiny]

@hhxdestiny thank you for reaching the OpenVINO!

Could you please check if the issue is present when you explicitly specify the device like below?

auto compiled_model = core.compile_model(model, "CPU"); 

Thank you for your help！The error is still the same! When I am not using an IR format model, such as using ONNX or PT format, the compile_model method works properly.

[andrei-kochin]

@hhxdestiny the error should not be the same as AUTO should no longer be used and there should be no error like:

[AUTO] compile model failed, CPU:Exception from src\inference\src\dev\plugin.cpp:54:
bad combination

Could you please double-check that you've hardcoded CPU device? Also would be worth to feed model path directly to the compile model like below:

auto compiled_model = core.compile_model("model.xml", "CPU"); 

[ilya-lavrenov]

The issue comes from xbyak, 7.06 is used in vcpkg, while OpenVINO repo uses older 6.69.1 version. Could you please try to set explicitly older version using pinning https://devblogs.microsoft.com/cppblog/take-control-of-your-vcpkg-dependencies-with-versioning-support/ ?

[hhxdestiny]

The issue comes from xbyak, 7.06 is used in vcpkg, while OpenVINO repo uses older 6.69.1 version. Could you please try to set explicitly older version using pinning https://devblogs.microsoft.com/cppblog/take-control-of-your-vcpkg-dependencies-with-versioning-support/ ?

Thank you for your suggestion. I will try it later. Does openvino have any plans to adapt to Xbyak 7.06 or 7.07?