Closed
Description
The current daemonset for k8s-neuron-device-plugin.yaml doesn't have any resource requests/limits. I take that setting requests/limits is a good strategy to prevent the container from hogging too many resources.
I'm thinking of something resembling this:
resources:
requests:
cpu: 200m
memory: 500Mi
limits:
memory: 500Mi
What I'm asking is: do the resource requirements change depending on the number of Inferentia chips available? Or depending on other unknown factors?
Metadata
Metadata
Assignees
Labels
No labels