Profiling with Tensorboard on Jetson TX2 without root

I got a Jetson TX2 with Jetpack 4.5 - Ubuntu 18.04 & ROS Melodic & ROS using Python 3.6.
Additionally I’m using a Virtual env for the cusignal environnement (with conda).

I wrote a rosnode in Python using cusignal and tensorflow library. Now I want to Profile some parts of the program with tensorboard.
I’m running into this issue: NVIDIA Development Tools Solutions - CUPTI_ERROR_INSUFFICIENT_PRIVILEGES: CUPTI Permission issue with Performance Counters | NVIDIA Developer
The Solution would be → run it with root.

Neither rosrun or conda virtual env recomend running with sudo. Is there another way to get access with the tensorboard profiler to the GPU?



Sorry that this is a limitation from the CUDA profiling.
You will need a root authority for this.