onnx2trt: No speed difference between models of different sizes.

[Fschoeller]

I have two yolov5 models of different sizes. One has 35.9m parameters, the other 12.7m. When I convert the models to TensorRT with trtexec --onnx=model.onnx --batch=5 --fp16 the resulting models have roughly the same inference speed (21 fps) even though the speed should be vastly different. What am I doing wrong?

[kevinch-nv]

What TRT version are you using? Are you able to provide the models you are benchmarking?

[Fschoeller]

I'm using TensorRT 7.1.3 on the Jetson Xavier AGX. Would you like the models as ONNX files?

[kevinch-nv]

Yes, proving the models in ONNX form will be useful.

Are you seeing the same performance difference with the latest version of TRT?