kubeflow/mpi-operator

Issues

• Add support for the managedBy field

  #646 opened 3 months ago by mimowo
  6

• How the file at tensorflow-benchmarks.yaml can run an MPI job ?

  #649 opened 2 months ago by luancaarvalho
  3

• What scale can mpi-operator support?

  #648 opened 2 months ago by yxzhao6
  3

• Worker pods not cleaned up upon `MPIJobEvicted` event

  #647 opened 3 months ago by shaowei-su
  0

• Question: Is the network traffic of AllReduce(like, ML gradients) encrypted between workers?

  #645 opened 4 months ago by jsyqrt
  10

• ttlSecondsAfterFinished for MPIJob, not only launcher

  #644 opened 5 months ago by hy00nc
  6

• "cleanPodPolicy: All" does not clean up launcher pod

  #643 opened 5 months ago by hy00nc
  1

• Connection reset

  #642 opened 6 months ago by bbenshab
  4

• how could mpijob of mpi operator worker get the hostname of launcher

  #641 opened 6 months ago by Oneal65
  2

• NCCL tests example

  #639 opened 6 months ago by samos123
  1

• Release 0.5.0

  #563 opened 7 months ago by tenzen-y
  11

• Running in a subset of namespaces

  #620 opened 9 months ago by emsixteeen
  8

• Wrong host info in discover_hosts.sh

  #621 opened 9 months ago by kuizhiqing
  0

• "make generate" command run failed

  #614 opened 9 months ago by wang-mask
  0

• The operator still creates the launcher when launcherCreationPolicy is "WaitForWorkersReady" and suspend is "true"

  #615 opened 9 months ago by wang-mask
  0

• MPI-Operator run example failed

  #598 opened a year ago by q443048756
  8

• Replace the plain pod workers with Indexed Job

  #613 opened 10 months ago by tenzen-y
  4

• Work with DeepSpeed for large scale training

  #611 opened 10 months ago by kuizhiqing
  28

• the object has been modified; please apply your changes to the latest version and try again

  #607 opened a year ago by gl-001
  8

• When when WaitForWorkersReady is enabled in MPI operator, MPI operator and gang scheduler are in a deadlock

  #608 opened a year ago by yzhao-2023
  4

• Cant get mpijob status when pod template is invalid

  #604 opened a year ago by congpeiqing
  9

• which is the latest mpi job definition between mpi-operator and training operator

  #605 opened a year ago by sxwl-donggang
  4

• OpenMPI 4.1.5

  #588 opened a year ago by bdevcich
  11

• Update stale examples

  #596 opened a year ago by jarulsamy
  2

• MPI Job fails on EKS with > 2 instances of 128 core per instance

  #594 opened a year ago by AymenFJA
  2

• Add support for linux/arm64 and linux/ppc64le for MPICH

  #565 opened a year ago by sheevy
  7

• Port conficts will occur when multiple pods dispatched to the same node under hostnetwork.

  #593 opened a year ago by Saturnoul
  3

• [feature]upgrade volcano to v1.8.0

  #586 opened a year ago by lowang-bh
  3

• pod priority was assigned to 0 though the priorityclassname of the podgroup had been assigned

  #592 opened a year ago by Robin7831
  2

• Add the ability to disable worker/launcher pod name suffix

  #591 opened a year ago by AymenFJA
  2

• Create CI pipeline for the exmaple images

  #541 opened 2 years ago by tenzen-y
  3

• Expected ssh contract that must be followed by images to use this operator

  #580 opened a year ago by aavbsouza
  3

• MPIJobs with Kubernetes Python SDK

  #582 opened a year ago by AymenFJA
  4

• Connection dropped after 24 hours

  #581 opened a year ago by sheevy
  2

• python setup.py doesn't appear to install?

  #578 opened a year ago by vsoch
  9

• Copy APIs from common repo into here

  #564 opened a year ago by tenzen-y
  3

• Add integration and e2e test for the volcano integration

  #568 opened a year ago by tenzen-y
  3

• Multiple MPI jobs via multiple launchers?

  #574 opened a year ago by AymenFJA
  6

• e2e test failed sometime

  #570 opened a year ago by lowang-bh
  4

• strange backup in hack/python-sdk/gen-sdk.sh

  #572 opened a year ago by lowang-bh
  2

• Consider handling the minResources when using volcano as a gang scheduler

  #535 opened a year ago by tenzen-y
  3

• how can i deploy distributed training on kubernete clusters with torch.distributed.launch

  #560 opened a year ago by ThomaswellY
  3

• questions about applying for nodes and gpus

  #558 opened a year ago by ThomaswellY
  9

• Pod scheduling conundrum

  #553 opened 2 years ago by sheevy
  8

• Seeing Launcher have two status: {'active': 1, 'failed': 1}

  #552 opened 2 years ago by kkolli
  2

• Running make scheduler-plugins-chart fails on second run

  #546 opened 2 years ago by alculquicondor
  3

• Add tolerations only to specific worker pods

  #539 opened 2 years ago by anxietymonger
  10

• Pass schedulingPolicy.PriorityClass to Launcher and Workers

  #536 opened 2 years ago by tenzen-y
  4

• Support exitCode restartPolicy

  #537 opened 2 years ago by Syulin7
  11

• Set the knowledge about Launcher and Worker to CRD

  #519 opened 2 years ago by tenzen-y
  3