[Bug]: core.import_model() fails to load compiled models after multiple export/import cycles

[jesspfjess]

OpenVINO Version

2025.2.0

Operating System

Other (Please specify in description)

Device used for inference

CPU

Framework

PyTorch

Model used

ResNet18

Issue description

Hi,
I wanted to do the following steps:

•       1. ✓ PyTorch model loaded
  

•       2. ✓ Converted to OpenVINO IR format
  

•       3. ✓ Loaded and compiled the converted model
  

•       4. ✓ Inference produces correct results
  

•       5. ✓ Exported compiled model 
  

•       6. ✓ Imported compiled model (no recompilation)
  

•       7. ✓ Re-exported compiled model to third directory
  

•       8. ✓ Re-imported compiled model from third directory
  

However, the model cannot be re-imported in the last step. Instead it is returning the error message that the model cannot be read

Step-by-step reproduction

I have the following code example which produces the iusse:

toy_example_pytorch_openvino.py

Relevant log output

2025-11-27T15:30:44.874+02:00:874 - pytorch_to_openvino_toy_example: INFO     Importing compiled model from /tmp/openvino_toy_example/third_export/ov_models/resnet18_compiled.bin...
2025-11-27T15:30:44.905+02:00:905 - pytorch_to_openvino_toy_example: ERROR    ✗ Failed to import compiled model: Exception from src/inference/src/cpp/core.cpp:189:
Exception from src/inference/src/dev/plugin.cpp:73:
Exception from src/inference/src/model_reader.cpp:223:
[ CORE ] Unable to read the model. Please check if the model format is supported and model is correct.



2025-11-27T15:30:44.905+02:00:905 - pytorch_to_openvino_toy_example: ERROR    Traceback (most recent call last):
  File "/home/algo-ipc/inspekto/inspekto/defect-predictor/src/defect_predictor/toy_example_pytorch_to_openvino.py", line 214, in import_compiled_model
    compiled_model = core.import_model(user_stream, "CPU")
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/venv/lib/python3.11/site-packages/openvino/_ov_api.py", line 668, in import_model
    super().import_model(
RuntimeError: Exception from src/inference/src/cpp/core.cpp:189:
Exception from src/inference/src/dev/plugin.cpp:73:
Exception from src/inference/src/model_reader.cpp:223:
[ CORE ] Unable to read the model. Please check if the model format is supported and model is correct.

Issue submission checklist

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

[YuChern-Intel]

Based on your shared codes, you are exporting the IR model (saved on disk) into compiled model for the first time. But for the second time, you are exporting the compiled model into another compiled model, which resulting in unsupported way of reading the model.

There is a more convenient API to compile the model in a single call, skipping the read step:

core = Core()
core.set_property({'CACHE_DIR': '/path/to/cache/dir'})
compiled_model = core.compile_model(model_path=xml_path, device_name=device_name)

[jesspfjess]

In my application, I need to export models and import them later, sometimes after recompiling them for specific use cases.
However, when I try to re-import an exported model that was previously imported, I get the error mentioned above.

The minimal steps to reproduce this issue are:

• export a compiled model
• import it
• export it again
• import it again <-- fails

[jgespino]

Hi @jesspfjess Could you please provide more insight into why you need to re-export a model that was already exported once? What is the use case? Once a model was exported once, you don't need to export it again, you can just import it with core.import_model().