hi our model in forward phase takes variable input size but in onnx conversion the input to the model must be fixed how can we solve this problem ? can we submit our model just by loading classes or models ...? or can we know about number of patient and voices to adjust our by them ?

Created by amin sefid aminreza_sefid
@aminreza_sefid I believe you can export the dynamic dimension for the input when you convert your model into ONNX format. Below example is in pytorch, you should get something similar in other frameworks as well. https://stackoverflow.com/questions/61850304/input-dimension-reshape-when-using-pytorch-model-with-coreml

onnx fixed input size page is loading…