I have a model trained over this network: https://github.com/facebookresearch/detectron2/blob/main/configs/COCO-Detection/faster_rcnn_R_50_FPN_3x.yaml
I want to run it natively on Jetson using TensorRT, however after exporting to ONNX and then trying to build an engine file I get an error.
TRT complains that TopK operator needs the K input to be a constant, which in this network is not.
I was thinking, this is a rather old model, 4 years old, and a rather popular one, so there must be a TensorRT version somewhere out there, or at least a well knows way for converting it to TRT.
Could you point me in the right direction how to run it natively?