Missing "libnvinfer-dev" after installing DRIVE OS 5.2 on DRIVE AGX

Please provide the following info (check/uncheck the boxes after creating this topic):
Software Version
DRIVE OS Linux 5.2.6
DRIVE OS Linux 5.2.0
DRIVE OS Linux 5.2.0 and DriveWorks 3.5
NVIDIA DRIVE™ Software 10.0 (Linux)
NVIDIA DRIVE™ Software 9.0 (Linux)
other DRIVE OS version
other

Target Operating System
Linux
QNX
other

Hardware Platform
NVIDIA DRIVE™ AGX Xavier DevKit (E3550)
NVIDIA DRIVE™ AGX Pegasus DevKit (E3550)
other

SDK Manager Version
1.6.0.8170
other

Host Machine Version
native Ubuntu 18.04
other

After installing and flashing my DRIVE AGX to have DRIVE OS 5.2 which include TensorRT 6.3, I noticed that TensorRT is not installed correctly.
When I run dpkg -l | grep TensorRT, I get

yet I am supposed to have something like the following

Dear @mohammed.el-yaagoubi,
It looks the installation is correct. Did not you see flashing is successful message with sdkmanager? I see the same output on my target with DRIVE OS 5.2.0

I understand that I need to do the cross compilation on my host machine first, however I have a problem while installing the files with SDKManager

Here is the debug logs

SDKM_logs_2021-09-06_19-09-20.zip (300.9 KB)

Dear @mohammed.el-yaagoubi,
Could you manually check installing TRT packages from the debians in sdkm download folder and let us know if it fixes the issue?

I just did and here’s the result. The problem is that I need TensorRT v6.3 with cuda 10.2

Dear @mohammed.el-yaagoubi,
Is the manual installation TRT 6.3 installation using debian is successful or you notice any errors? Do you see any folder like /var/nv-tensorrt-repo-cuda10.2-trt6.3.1.3-x86-host-ga-xxx on host?

Yes I can see this folder. What I don’t understand is that I have now 2 versions of Cuda and Tensor RT

Dear @mohammed.el-yaagoubi,
If you want TRT 6.3 packages to installed on host, could you go to /var/nv-tensorrt-repo-cuda10.2-trt6.3.1.3-x86-host-ga-20200724 and force dpkg to install the needed package like sudo dpkg -i --force-overwrite <debian package> and see the reflected package in dpkg -l to confirm. Once the TRT issue is fixed, if you check installing DRIVE OS 5.2.0 via sdkmanager, I would expect to have smooth installation/flashing. Let us know how it goes.

What I need is to install Tensor RT v6 on the target (AGX DRIVE) via cross compilation

Dear @mohammed.el-yaagoubi,
Could you check the deb files in /var/nv-tensorrt-repo-cuda10.2-trt6.3.1.3-d5l-cross-ga-20200724 as well. I expect /usr/src/tensorrt folder should have samples and you can cross compile the samples for target. I could see in the logs that CUDA/CUDNN packages are installed on host which out any issue.

Note that, flashing is required to install TRT libs on target. I assume you already have DRIVE OS 5.2.0 on target and looking for setting up cross compilation on host.

I have already installed and flashed the AGX DRIVE to have DRIVE OS 5.2. I don’t want to repeat this step another time. However I noticed that Tensor RT v6 is not installed on the target. That’s why I would like to know how to do a cross compilation to install Tensor RT v6.3 on AGX DRIVE.

Dear @mohammed.el-yaagoubi,
However I noticed that Tensor RT v6 is not installed on the target.

How did you verify target has no TRT 6.3? Could you confirm did you see flashing is successful message when you flashed the target? If not, flashing is failed.

how to do a cross compilation to install Tensor RT v6.3 on AGX DRIVE

Cross compilation is needed to compile TRT sample to run on target. It does not install TRT 6.3 on target. You need to flash the target using sdkmanager to update DRIVE OS SW