Converted pytorch model doesn't work

hey guys as you can see in the image I converted a pytorch model to onnx, then to coreML but my app crashes once I launch it on my phone, the problem has to do with the input and output names , cormel has input referring to an image and output referring to a string, idk how to implement that when converting from pytorch, any help would be much appreciated.


Converting quantized model to ONNX, isn’t supported fully yet. I believe it’s an experimental feature??

It could be a few things, but I’d have suggest you test the PyTorch, then ONNX, and finally the CoreML. Pin down the problem.

Can you share the Jupiter notebook on the left?

I want to try it to make sure it’s not an issue related to the size of your input dummy variable.