Tensorrt and cudnn on drive os 6.0.8

Please provide the following info (tick the boxes after creating this topic):
Software Version
[-] DRIVE OS 6.0.8.1
DRIVE OS 6.0.6
DRIVE OS 6.0.5
DRIVE OS 6.0.4 (rev. 1)
DRIVE OS 6.0.4 SDK
other

Target Operating System
[-] Linux
QNX
other

Hardware Platform
DRIVE AGX Orin Developer Kit (940-63710-0010-300)
DRIVE AGX Orin Developer Kit (940-63710-0010-200)
DRIVE AGX Orin Developer Kit (940-63710-0010-100)
DRIVE AGX Orin Developer Kit (940-63710-0010-D00)
DRIVE AGX Orin Developer Kit (940-63710-0010-C00)
[-] DRIVE AGX Orin Developer Kit (not sure its number)
other

SDK Manager Version
1.9.3.10904
[-] other

Host Machine Version
native Ubuntu Linux 20.04 Host installed with SDK Manager
[-] native Ubuntu Linux 20.04 Host installed with DRIVE OS Docker Containers
native Ubuntu Linux 18.04 Host installed with DRIVE OS Docker Containers
other

Hi, I have some problems after installing 6.0.8 with docker. I only see the bin folder in usr/local/tensorrt, making tensorrt unavailable. The deb package I pulled from the image is the amd64 version which is also not installable and cudnn is not available, what should I do to fix it?

Dear @zhangpeixing,
If you are looking for TensorRT samples, you can copy them from docker and check compiling on target. The TRT/cudnn libraries should be part of target already.

How do I do that, in the docker?

Dear @zhangpeixing,
copy the samples folder from /usr/src/tensorrt on docker to target at /usr/src/tensorrt/ and check compiling on target directly.

Is it copied to orin’s home/nvidia/tensorrt?

Dear @zhangpeixing,
Yes. The sample folder should be copied to target at /usr/src/tensorrt/

I’m using tensorrt for inference and I can’t find the nvinfer.h file. And I found no include and lib files under usr/src/tensorrt file. I don’t know much about this, can you explain me in detail?

Dear @zhangpeixing,
For TensorRT sample cross compilation, please check Build and Run Sample Applications for DRIVE OS 6.x Linux | NVIDIA Docs

If you want to compile the samples on target, The TRT libs can be found at /usr/lib/aarch64-linux-gnu on target . TRT headers are not part of target. The headers can be found at /usr/include/aarch64-linux-gnu on docker . You need to copy the missing header files to /usr/include/aarch64-linux-gnu on target. Copy the TRT samples to /usr/src/tensorrt/ on target.
For sample compilation check running make from /usr/src/tensorrt/samples/ folder. Let us know if you notice any issues.

Hello, I still get the following errors after compiling.

./../bin/dchobj/sampleAlgorithmSelector/sampleAlgorithmSelector/../common/logger.o ../../bin/dchobj/sampleAlgorithmSelector/sampleAlgorithmSelector/../common/getOptions.o ../../bin/dchobj/sampleAlgorithmSelector/sampleAlgorithmSelector/../common/sampleReporting.o ../../bin/dchobj/sampleAlgorithmSelector/sampleAlgorithmSelector/../common/sampleEngines.o -Wl,--end-group
/usr/bin/ld: cannot find -lnvinfer
/usr/bin/ld: cannot find -lnvparsers
/usr/bin/ld: cannot find -lnvinfer_plugin
/usr/bin/ld: cannot find -lnvonnxparser
/usr/bin/ld: cannot find -lcudnn
collect2: error: ld returned 1 exit status
make[1]: *** [../Makefile.config:363: ../../bin/sample_algorithm_selector_debug] Error 1
make[1]: Leaving directory '/usr/src/tensorrt/samples/sampleAlgorithmSelector'
make: *** [Makefile:83: all] Error 2

I found that my Orin system’s TensorRT does not have the lib and include files. What should I do?

I would greatly appreciate your help in solving this issue. Thank you very much.

Dear @zhangpeixing,
Did you check cross compilation steps on docker? I expect it to work out of the box.

I used make TARGET=aarch64 on orin without success, prompting me that I was missing the file in the image. Compiling inside docker and cp to orin after successful compilation also doesn’t work. And there are only bin, data, samles in the tensorrt folder is that correct?

Dear @zhangpeixing,
As clarified in Build and Run Sample Applications for DRIVE OS 6.x Linux | NVIDIA Docs
You need to run sudo make TARGET=aarch64 on docker container on host.

I have tried the above steps and copied the compiled files from the Docker environment to /usr/src/tensorrt on the Orin device. What should I do next?

image
Is it enough to copy the entire tensorrt file, are there any other dependencies that need to be copied?

Dear @zhangpeixing ,
Please copy the tensorrt folder to /home/nvidia and check running samples. Please see Build and Run Sample Applications for DRIVE OS 6.x Linux | NVIDIA Docs . Let us know if you see any runtime issues.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.