It seems that after deploying the gpumanager project, there is an issue with controlling GPU memory. I have set the pod quotas as follows, but the control over GPU memory and computing power does not take effect. What's even more peculiar is that if I wait for the same pod for over 60 minutes, the restrictions are likely to take effect. There are no obvious errors in the gpumanager logs.
resources:
limits:
tencent.com/vcuda-core: "50"
tencent.com/vcuda-memory: "32"
requests:
tencent.com/vcuda-core: "50"
tencent.com/vcuda-memory: "32"