When there are multiple inputs rather than only an image input, how should the calibration dataset be constructed?

Maybe this helps:

Hailo Community - How can i optimize the model which have multi input