Troubleshooting

The default port for running OMNI Agent web server is busy (Windows)

By default, OMNI Agent runs on port 8080. If this port is busy, the web server becomes unavailable.

Solution: follow the link

OMNI Agent installer does not run on some Linux (Ubuntu) distributions

During installation, two error windows appear - one of them with a message about an unsuccessful authorization attempt.

Solution: copy the command from the error window to a terminal, run it with sudo (or as a user with root privileges) and click "OK" on the error window.

OMNI Agent does not start when GPU is enabled

Solution: Follow the steps below to check your graphics card:

1. Check the CUDA Toolkit installation and make sure that the NVIDIA drivers are suitable for the installed version.

   1.1 To find out the version of CUDA Toolkit installed on your computer, use the command:

   nvcc --version

   In the example below, we have CUDA Toolkit 11.8 installed on our computer.

   

   Supported CUDA Toolkit versions. If you need to update or install a new version, follow the instructions on the NVIDIA website.

   1.2 Check CuDNN version:

   Linux:

   cat /usr/local/cuda/include/cudnn_version.h | grep CUDNN_MAJOR -A 2

   Windows:

   cat "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDNN8.5\include\cudnn_version.h” | grep CUDNN_MAJOR -A 2

   1.3 Make sure the latest NVIDIA driver is installed and working correctly:

   nvidia-smi

   The command output displays the version of the installed NVIDIA driver and the maximum CUDA Toolkit version that this driver version supports.

   In the example below, NVIDIA driver version 530.41.03 supports CUDA Toolkit version 12.1 or lower.

   

   You can also update the drivers on your computer and disable auto-update of drivers. To check the compatibility of the CUDA Toolkit version with drivers, follow NVIDIA website.

2. To check the overall functionality of CUDA Toolkit, compile and run CUDA Samples.