Abhishek Malvankar comments

Results 77 comments of


                                            Abhishek Malvankar

New strategy for MIG enable GPUs that will advertise no MIG slices if not created

Thank you for creating the fix for the mixed strategy. we are facing issues using MIG Manager: - Start MIG manager with `mig.config` as `all-enabled` - Let the operator configure...

New strategy for MIG enable GPUs that will advertise no MIG slices if not created

Thanks @empovit. @klueska @cdesiniotis any pointers that can help us with step 6 of the above comment?

Add resource check before dispatching workloads

> Btw, in general, quota management is already a good approximation for what is available in a cluster and people are using Kueue in non-autoscaled environments in production just based...

Add resource check before dispatching workloads

> Since you are coming from DRA, I think we can make this feature request about DRA. Then, what we would need is a quota model for DRA resources. Not...

Add resource check before dispatching workloads

> FYI: @asm582 > > I have a similar situation in the required multiple GPUs per pods. We can imagine the following situation: > > * The job has 2...

[Discussion] Consider pytorchJob cleanPodPolicy in kueue

"A better way seems like we should reclaim the completed pods rather than the whole job, but the tricky thing is it's hard for the kueue to aware of the...

Allow configuring replicas for raw deployments

> Isn't it related with #4297 ? This PR fixes the deployment reconcile, fixes the replca count too. Are you trying to answer this question? https://github.com/kserve/kserve/pull/3759#discussion_r1987345380