Collect gpu diagnostics informations from NVML and repulish them as ROS diagnostics. Multi-gpu is supported.
On Ubuntu you will need to install nvml as cuda packages cannot be made part of rosdeps.
apt-get update
apt-get install cuda-nvml-dev-YOUR-CUDA-VERSION
- Launch the node with the provided example launch file
roslaunch ros_nvml_diagnostics gpu_diagnostics.launch
- Viewing the data
rosrun rqt_runtime_monitor rqt_runtime_monitor