TensorRT Cask Error in checkCaskExecError<false>

[smaiah.sarah]

Environment

**Nvidia card: Jetson AGX Xavier
TensorRT Version: 8.0.1.6
CUDA Version: 10.2.300
CUDNN Version: 8.2.1.32
Operating System + Version: Ubuntu 18.04

Do you have any idea how to solve this problem
Thanks

[SivaRamaKrishnaNV]

Dear @smaiah.sarah,
Is the engine generated on same machine with same TensoRT version?

[smaiah.sarah]

Yes, the engine was generated on the same machine with same TensorRT version

[SivaRamaKrishnaNV]

Dear @smaiah.sarah,
Could you please share the reproducible steps and relavant files

[smaiah.sarah]

I used the scripts in the following repository to convert my yolo model to TRT:
https://github.com/jkjung-avt/tensorrt_demos/tree/master/yolo

and the following repository for object detection detection
https://github.com/indra4837/yolov4_trt_ros

The error is with the following script
trt_yolo_v4.py (5.9 KB)

The previous script calls the following script for the inference
yolo_with_plugins.py (12.3 KB)

[smaiah.sarah]

Dear @SivaRamaKrishnaNV
Do you have any idea about thIs error?
Is it a metter of multithreading ? the TRT and the inference in different Threads?

[SivaRamaKrishnaNV]

Dear @smaiah.sarah,
The error indicate issue with CUDA context.
Is it possible to check loading engine with trtexec to confirm no issue with engine.

[smaiah.sarah]

Dear @SivaRamaKrishnaNV
I tried to launch trtexec and i get th following error

[SivaRamaKrishnaNV]

Dear @smaiah.sarah,
The error in trtexec indicates, it has a non supported layers which are implemented like plugins.
Are you running the it like ROS node?
I would recommend to get the TRT model preparation/inference code from repo and test on target to see if it working correctly with out any issues.