Python API vs C++ API

[fadoughou]

Description

This is not a bug post but rather a general question. I am seeking advice.

My team has been developping AI pipelines for Jetson devices. We use c++ API (e.g. JetPack 4.6, TRT 8.0).
We are considering moving to python API because python offers more flexibility to build pipelines faster and process data.
However, I would like to understand:

1. Is there any limitation of python API compared to c++ API.
2. Is python API supported with latest releases of JetPack and TRT?
3. How much performance degradation comparing to c++ API? Especially, we aim to offer a pipeline with the most optimized fps.

Environment

TRT 8.0, 8.2:
GPU Jetson nano, Jetson AGX, Jetson Orin NX:
PyTorch Version (3.x):

[AakankshaS]

Hi @fadoughou ,
For the post you may get better support on Jetson Forum.

Moving.
thanks

[AastaLLL]

Hi,

1. Python can work on Jetson normally.
2. On JetPack, it usually installs both C++ and Python API.
3. For the same TensorRT function, the performance is similar.

Thanks.