__forward__impl meaning?

Hi, Could someone explain why shufflenetV2 code in torchvision need to use __forward__impl?

The explanation in the code say that I can check with TorchScript super().

But I have checked with TorchScript tutorial and have not found that why we need to use __forward__impl().

Is that coupled with TorchScript and can speed up the inference when model is being deployed?

1 Like

I found more explanation here.

1 Like