DriveWorks Samples in DriveOS 7.0.3 docker container are nto running due to CUDA Version mismatch

Autonomous Vehicles DRIVE AGX Thor DRIVE AGX Thor General
1

DRIVE OS Version: Nvidia DriveWorks 7.0.3.0-0010 (tag nvcr.io/drive/driveos-sdk/drive-agx-linux-nsr-aarch64-sdk-build-x86:7.0.3.0-0010)

**Issue Description: NVCC, Cuda Runtime Not configured in Drive 7.0.3 Container (**drive-agx-linux-nsr-aarch64-sdk-build-x86)

I’m using Nvidia Drive AGX Thor with DriveOS. I followed the step from DriveOS 7.0.3 Installation Guide using docker container to install DriveOS on host pc.

I wanted to run the samples on the host PC running the container.

I am not able to run any sample on the DriveOS container. The following seem to be the missing links:

  1. nvcc is missing in the container. Cannot compile Cuda code in the container.

  2. The following error shows up: /usr/local/driveworks/targets/x86_64-Linux/lib/libdw_base.so: undefined reference to `cudaGraphExecDestroy@libcudart.so.11.0’

    1. indicating the wrong Cuda runtime version configuration mismatch between the AGX thor and this container.

Can you please provide feedback on following questions:

  1. How can i fix the Cuda Runtime version mismatch issue ?
  2. if NVCC is not part of the container, then how do i build cuda applications on this container ?
  3. Is there any conatiner that Nvidia can provide that contain all these dependencies to build complete applications for the Thor?
  4. Do you have any release with the bug fixes for this contianer version (7.0.3.0-0010)?
4

Dear @keshav.poygaipakkamkhanda ,
It is noticed from DOS 7.0
3 release notes that, tools, samples are not supported on x86.
Please see DriveWorks tools in DriveOS 7.0.3 docker fail to run due to CUDA version issue - #6 by SivaRamaKrishnaNV

5

Could you let me know when the next release is happening ? Is there a workaround to continue developing applications.