How to compare the inference speed before and after conversion

[ChenZhenGui]

I would like to know how to compare the inference speed of the pytorch model and the model after converting to tensorrt

[lvhan028]

Please try tools/profiler.py

[ChenZhenGui]

so what's the difference between test.py and profiler.py?

[lvhan028]

profiler.py is especially for speed test.
The main purpose of test.py is to check the metrics consistency between pytorch models and backend models