Onnx python post-processing vs. TAO train post processing

I added the above parameter to my export API call and it fixed the issue.

Can this be added to the documenation here, DetectNet_v2 - NVIDIA Docs?

I didn’t see any mention of the onnx_route parameter. Maybe it is explaned somewhere else but it could have saved me a lot of time if it was in here. But anyway, thank you!