How can I share weights between two models loading on same device?

[Farhadi]

Hello,

Is it possible to share weights between two models running on the same device?

Also, can we change the model input size after loading it in the device (not before loading same as "Shape Inference")? I want to change the model input size while it's running.

Thanks,

[Iffa_Intel]

Greetings,

You may refer to this: https://www.youtube.com/watch?v=Ga8j0lgi-OQ

This is about utilizing a model optimizer to adjust the model size

Sincerely,

Iffa

[Farhadi]

Thanks,

I am looking to reshape the model while the model is running on the device, not before loading on the device. I follow the provided instruction (https://docs.openvinotoolkit.org/latest/openvino_docs_IE_DG_ShapeInference.html), but it is for prior loading the model on the device. Actually, I found out there is no way to reshape the model after loading the device.

I also need to load several models on the device, just the weights for these models are same, but it seems OpenVino doesn't support weight sharing.     

[Iffa_Intel · Accepted solution]

The page https://docs.openvinotoolkit.org/latest/openvino_docs_IE_DG_ShapeInference.html (the ones that you are referring to) provides all available methods for runtime model reshaping. We don't have options of model reshaping other than mentioned there as of now.

Also, regarding your question for weights sharing, model weights are stored within .bin file of the resulted IR model. We don't have an embedded option to implement this, however, you could simply use multiple instances of the same .bin file.

Sincerely,

Iffa

[Iffa_Intel]

Greetings,

Intel will no longer monitor this thread since this issue has been resolved. If you need any additional information from Intel, please submit a new question. 

Sincerely,

Iffa