CoreML not using GPU

[Jasonsey]

Hi there,

I'm running an onnx model useing CoreML. if I run this model in cargo I can see it's using gpu well, it spends about 40s to infer the model.

But If I build this model into Mac APP using cargo swfit crate, only cpu is used and it spends about 2min to infer the model. And also I try to infer the model using python onnxruntime, it spend about 10s.

this is a snapshot when I run this model in Mac APP.

Any Idea that I can find why this happen or I can I solve this issue ?