Use onnx function only in tracing mode

[prabhat00155]

#5310 introduced extra operations that may affect performance. Adding conditional checks to only use onnx functions when in tracing mode.

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit 7ca71f0 (more details on the Dr. CI page):

---

• 1/1 failures introduced in this PR

---

1 failure not recognized by patterns:

Job	Step	Action
cmake_macos_cpu	curl -o conda.sh https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-x86_64.sh
sh conda.sh -b
source $HOME/miniconda3/bin/activate
conda install -yq conda-build cmake
packaging/build_cmake.sh
	🔁 rerun

---

This comment was automatically generated by Dr. CI (expand for details).

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

[zhiqwang]

Hi @datumbox and @prabhat00155 , do you think that we should also add a unit-test for SSD and fcos when exporting to ONNX, as I know, the downstream TVM also use the JIT tracing mode.

[datumbox]

@zhiqwang If you have the capacity to do this on a follow up PR that would be lovely. One issue we faced previously with onnx unit-tests is that they are very brittle and easy to break. I personally don't have much experience working with ONNX so if you can do it, I'll take it 😄

[zhiqwang]

Hi @datumbox

If you have the capacity to do this on a follow up PR that would be lovely.

I can try to add it! Recently I also found that TensorRT 8.2+ has very good support for ONNX. It will become very easy for one stage models (SSD / RetineNet / Fcos) to be deployed on TensorRT through ONNX.