Please help me

My virtual machine is 16.04. I used SDKmanager to install drivers for JETSON AGX. It installed the cuda toolkit for l4t 10.0, but stopped at 95.5%. why?

02:03:05 INFO : CUDA Toolkit for L4T : exec_command: scp -F /dev/null -o PubkeyAuthentication=no -o UserKnownHostsFile=/dev/null -o StrictHostKeyChecking=no -o ConnectTimeout=5 -r /home/mxq/Downloads/nvidia/sdkm_downloads/cuda-repo-l4t-10-0-local-10.0.166_1.0-1_arm64.deb mxq@
02:03:05 WARNING : CUDA Toolkit for L4T : Warning: Permanently added ‘’ (ECDSA) to the list of known hosts.
02:03:05 INFO : CUDA Toolkit for L4T : [target] mxq@’s password:
02:03:05 INFO : CUDA Toolkit for L4T :

please help me.

Virtual machines are not supported. They tend to lose connection on USB or ethernet. On the other hand, the location where things stop makes me wonder if the first login has not yet occurred (which would not be a VM issue), and thus the login account/password wouldn’t work. Not sure from that little bit of log. If you just boot the Xavier normally, do you have the ability to log in and reboot? Is “mxq” the account on the host, or on the Xavier?

But I successfully performed the previous operations and connections with the virtual machine. I am able to start XAVIER normally. Mxq is the account I created earlier.

You would probably need to start the Xavier normally, and then run SDK Manager with only the CUDA/CUDA Toolkit part selected (don’t flash), and then provide the full log for the install to see which parts work and which don’t. It’s hard to tell without more logging. Also, before starting, be sure to mention how much space the VM has at the location of running SDKM (“df -H -T ~/nvidia” would give good information).

During the package install run ping to the same IP address that SDKM is using for package transfer. Normally I think that is “”, so something like “ping” in a separate terminal, and note if ping still works during the stall. If you use a different IP address, then adjust for that.