TensorRT Docker:: NVIDIA GPU Quadra Series P2000:: PC reboot issues after installation of drivers

[vyom.mishra]

Dear Team,

I am trying to set up a docker container for TensorRT for host Linux and for Jetson cross-compilation. Following this guide(GitHub - NVIDIA/TensorRT: TensorRT is a C++ library for high performance inference on NVIDIA GPUs and deep learning accelerators.)

I have set up the docker containers successfully(Install Docker Desktop on Ubuntu | Docker Documentation). When I tried to create a container accessing all_gpu(command: ./docker/launch.sh --tag tensorrt-ubuntu18.04-cuda11.4 --gpus all) it shows the NVIDIA driver accessing GPUs issue.

Then I freshly downloaded the drivers and installed them successfully on the Linux host(Linux x64 (AMD64/EM64T) Display Driver | 510.73.05 | Linux 64-bit | NVIDIA)
Drivers found are below:

lspci | grep -i nvidia

01:00.0 VGA compatible controller: NVIDIA Corporation GP107GLM [Quadro P2000 Mobile] (rev a1)
01:00.1 Audio device: NVIDIA Corporation GP107GL High Definition Audio Controller (rev a1)

, while restarting the PC, the boot got stuck in between.

I have followed all pre-requisites.

Can you please let me know the root cause for the same?

Thanks and Regards,
Vyom Mishra

[spolisetty]

Hi,

To install TensorRT, please refer installation guide.

Installation Guide :: NVIDIA Deep Learning TensorRT Documentation

Also please refer support matrix doc and make sure, dependencies are installed.

Support Matrix :: NVIDIA Deep Learning TensorRT Documentation

You can also use the TensorRT NGC docker container to avoid setup-related issues.
https://ngc.nvidia.com/containers/nvidia:tensorrt

[vyom.mishra]

Dear spolietty,

I have not faced any issues in docker and tensorRT setup.

When I installed the NVIDIA drivers needed as a pre-requisite to install the Cuda toolkit, it got successfully installed but when I tried to reboot the Laptop for changes it didn’t boot.

Kindly suggest me the solution to resolve the issues related to NVIDIA drivers to access GPUs.

And kindly let me know the difference between both docker and NGC docker for running any deep learning model? Will NGC docker doesn’t need NVIDIA drivers for accessing the GPUs?

I will try to follow the link shared by you and try to run the NGC docker for model inference.

Thanks and Regards,
Vyom Mishra

[spolisetty]

NGC containers are GPU-optimized containers such that running on Nvidia GPU with Nvidia container should have best possible performance. NGC containers as well need drivers.

This issue looks like more related to setup related issue. We are moving this post to a relevant forum to get better help.

Thank you.

[vyom.mishra]

Dear Team,

I am able to resolve the issue.

Thanks for your support.

Regards,
Vyom Mishra