FP16 model unable to run in OpenVINO

[DarkHorse]

Hi,

My customer has train their AI model using PyTorch framework using Intel Sonoma Creek and they want to run the AI Inference on a VPU accelerator card. 

They managed to convert their model from:

PyTorch -> ONNX -> FP16 IR Format

I have confirmed their FP16 model on my OpenVINO 2021 and showing similar error messages:

[ INFO ] Parsing input parameters
[ INFO ] Reading input
[ INFO ] Loading Inference Engine
[ INFO ] Device info:
[ INFO ] CPU
MKLDNNPlugin version ......... 2021.4
Build ........... 0
Loading network files
[ ERROR ] Unknown model format! Cannot find reader for model format: xml and read the model: model_barehand.xml. Please check that reader library exists in your PATH.

Customer have no issue running FP32 model in OpenVINO 2022 on CPU.

Customer also tried FP16 model person-detection retail 002 from intel it working ok

Can someone have a look into this?

Thanks.

[Peh_Intel]

Hi DarkHorse,

For your information, in the latest version of OpenVINO™, the error message has changed... earlier it explicitly reported that the found IR-version is not supported... now it fails with "Unknown model format! Cannot find reader for model format".

In short words, OpenVINO™ 2022 is able to load IRv10 (converted by Model Optimizer 2021.4) and IRv11 (converted by Model Optimizer 2022.1) into Inference Engine whereas OpenVINO™ 2021.4 is only able to load IRv10 and will produce the above error if loading IRv11.

Regards,

Peh

[DarkHorse]

Hello Peh,

My customer is seeing the same error messages even he is trying to run the inferencing in OpenVINO 2022?

Attached is the log files... can you have a look?

Thanks.

[Wan_Intel]

Hi DarkHorse,

Referring to this article, could you please check if both inference_engine_ir_reader.dll and inference_engine.dll are located in the following directory:

<INSTALL_DIR>\deployment_tools\inference_engine\bin\intel64\Release

Regards,

Wan

[Peh_Intel]

Hi DarkHorse,

Thank you for your question. If you need any additional information from Intel, please submit a new question as this thread is no longer being monitored.

Regards,

Peh

[Farhâd]

I see this same problem on the raspberry pi4b.  Please let me know the solution.