When I use ./webodm.sh start --gpu I get at most 3/5 of the webodm containers running under container view in docker. At times I get none running. The nodeodm.gpu container never runs.
I am still able to open the webodm ui though. Just after starting to processing, docker always exits. The process will continue to finish. I get “nvidia-smi detected, using GPU for sift extraction” in the process log. I see GPU usage when I run qnvsm.
When I run ./webodm.sh start, I get 5/5 containers running. Docker doesn’t exit and the process proceeds to completion. The most strange thing is that it processes faster than ./webodm.sh start --gpu.
Also after adding user, when i run nvidia-smi i get:
NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running
I managed to get 5/5 containers running when I run ./webodm.sh start --gpu.
I replaced docker desktop with docker engine 20.10.17.
Running webodm with --gpu now processes faster than with CPU only.
The problem is that the gpu process is just about 3mins faster than CPU on 1000 images. I need help to fix that.