I have a single node deployment of AICenter. After training and generating a skill through the pipeline in AICenter, my memory usage increases by 2GB. Even if I don’t call this skill, the 2GB of memory is not released, which means I don’t have enough memory to train another skill. Why does the generated skill occupy memory? Shouldn’t it only occupy memory when called? This means that if I don’t delete old skills, I can’t train new ones.
My Automation Suite version is 22.10.2, and it is deployed as a single-node local deployment.
Before generating the skill, the computer memory usage like below:
total used free shared buff/cache available
Mem: 62 50 4 0 7 11
Swap: 9 7 2
after generating 1 skill, the computer memory usage like below:
total used free shared buff/cache available
Mem: 62 51 2 0 8 10
Swap: 9 8 1
So Mem of free increases by 2GB，Why does the generated skill occupy memory? Shouldn’t it only occupy memory when called?
I’ve noticed that whenever the free memory on my Linux server runs low, I can’t train new skills through AICenter.
Hi @pjj1618
Welcome to UiPath community.
We are also having similar deployment as you are having. In standalone deployment, by default when you deploy a ML skill it consumes memory and CPU to make it available all the time. it is not that it occupies memory only when we call the skill, it occupies it all the time when it is deployed and made available.
if you does not have enough memory to train another model, either you have to increase your memory or stop the currently deployed ML skill so that it releases the memory it occupied.
Hope this helps.
Cheers