TensorRT inference context in ROS callback

[ec2020]

I’m noticing this error message has been reported elsewhere:

TensorRT 6.0 error

I have been using TensorRT 5.0 with my app without a problem, but after upgrading to the latest version 6.0 on Ubuntu 18.04 with RTX gpu, I sometimes get the following error message: FAILED_EXECUTION: std::exception ../rtSafe/safeContext.cpp (133) - Cudnn Error in configure: 7 (CUDNN_STATUS_MAPPING_ERROR) The only change in the setup the upgrading of tensorrt so I feel like this is a problem brought by the upgrade. Do you have any idea why this error occurs ?

Jetson Nano TensorRT heisenbug

Hi, Thanks for the sharing. We are trying to reproduce this issue internally. (multiple client connection) Will let you know about the following.

I am getting this error when I have tensorRT and another pycuda program both accessing the GPU (one after the other) in a ROS callback like documented here