Survey on usage TVM based packages

[lexavtanke]

There are two main inference engines for ml models in Autoware: TensorRT and Apache TVM.

From my perspective only TensorRT is mainly in use.
As for now there are some changes in TVM packages

• build(tvm_utility): remove download logic from CMake and update documentation autoware.universe#4923
• build(lidar_centerpoint_tvm): remove artifacts download autoware.universe#5367
• build(lidar_apollo_segmentation_tvm and lidar_apollo_segmentation_tvm_nodes): remove download from cmake autoware.universe#5431

These changes mainly do not affect current usage of the packages. But there is some proposal about updating  tvm_utlity package which allows to dynamically load model during the inference. It will allow to use TVM models with more ease.  

So I'm curious about how many people are interested in usage of the Apache TVM?
Because It make sense to do this change only if there is demand from the community.