CUDA Installation failed on SDK Manager 2.1.0 for Jetson AGX Xavier

[learning_23]

I am currently trying to flash Jetson AGX Xavier. Below are my systems:

Ubuntu 20.04

SDK Manager 2.1.0

JetPack 5.1.2

Settings:

OEM Configuration: Pre-Config

Storage Device: NVMe (using a SSD that I have installed on the agx xavier)

Connected by USB C to USB 3 port, and ethernet cable is also connected to my PC.

Network: I have created a profile that is 192.168.55.5 netmask 255.255.255.0

For the Connection: USB

When I tried using the default ip of 192.168.55.1, it has the error that the default ip is unavailable. However, I am able to ping it. I tried to use the IP address of the host (later on I found that this is incorrect) and it was stuck on installation with Error for the CUDA installations.

Does anyone know how to resolve these issues?

Thank you!

[AastaLLL]

Hi,

Please help to share the error message with us.

Thanks.

[learning_23]

Hello! Here’s the screenshot of the error message I have. I was only able to bypass this after I key in the host pc’s ip address instead of the agx xavier.

Afterwhich, I was stuck here:

This is the zip file of the exported debug log
SDKM_logs_2024-07-03_13-45-41.zip (787.0 KB)

[AastaLLL]

Hi,

13:41:38.834 - error: NV_L4T_CUDA_RUNTIME_TARGET_POST_INSTALL_COMP@JETSON_AGX_XAVIER_TARGETS: dpkg: error processing archive /opt/nvidia/deb_repos/cuda-repo-l4t-11-4-local_11.4.19-1_arm64.deb (--unpack):
13:41:38.834 - info: Event: NV_L4T_CUDA_RUNTIME_TARGET_POST_INSTALL_COMP@JETSON_AGX_XAVIER_TARGETS - error is: dpkg: error processing archive /opt/nvidia/deb_repos/cuda-repo-l4t-11-4-local_11.4.19-1_arm64.deb (--unpack):
13:41:38.834 - info: NV_L4T_CUDA_RUNTIME_TARGET_POST_INSTALL_COMP@JETSON_AGX_XAVIER_TARGETS:  package architecture (arm64) does not match system (amd64)

Are you using the Xavier IP or host server IP?
Please maks sure you are using the XavierIP.
If the local IP is not working, could you connect Xavier to the network and try it again?

Thanks.