Trouble with xavier-nx 16gb emmc

Autonomous Machines Jetson & Embedded Systems Jetson Xavier NX
1

Hello.

I have troubles with cuda on my xavier-nx 16gb emmc jetpack version 5.1.2 (only system without cuda from sdk image) :

  1. I installed cuda with command

sudo apt-get install nvidia-cuda

in .bashrci add PATH="/usr/local/cuda/bin:$PATH" and LD_LIBRARY_PATH=“/usr/local/cuda/lib64:$LD_LIBRARY_PATH”

But then i try used nvcc i get this message: bash: nvcc: command not found
2. Me also need tensorrt sudo apt install nvidia-tensorrt ? or something exotic ?
I guess i miss something.

3

Hi,

Suppose CUDA can be installed with the below command:

$ sudo apt-get install nvidia-cuda nvidia-cuda-dev

After installation, please check if /usr/local/cuda/bin/nvcc exists.
TensorRT follows the similar instruction:

$ sudo apt-get install nvidia-tensorrt nvidia-tensorrt-dev

Thanks.

1 Like
4

Got it. And it works! Thank you!

But next question, so as you know i have limit memory only 16 gb . And i have already deleted by this tutorial: link So nvidia-cuda-dev and tensorrt-dev are necessary package to compile onxx -> engine or it is just plugin with addition function ?

5

Hi,

To compile a TensorRT engine, you will need to deploy a possible algorithm (use CUDA, cuBLS, cuDNN, …) on the target.
Although TensorRT supports inference without cuDNN, CUDA is required for conversion and inference.

Thanks.

8

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