kubeflow/trainer

Enhanced MPI Support in TrainJob

Fermée

#2 751 ouverte le 24 juil. 2025

 (26 commentaires) (3 réactions) (0 personne assignée)Go (964 forks)github user discovery
area/controllerarea/mpiarea/runtimesgood first issuehelp wantedkind/feature

Métriques du dépôt

Stars
 (2 112 étoiles)
Métriques de merge PR
 (Merge moyen 10j 9h) (75 PRs mergées en 30 j)

Description

What you would like to be added?

We plan to enhance MPI support in TrainJob in these areas:

  • SSH-based node communication
  • Add resource per node support
  • Address the failing launcher issue in the initialization phase (launcher fails until the MPI worker is ready)

Why is this needed?

As we discussed in the WG call yesterday, this will be included in Kubeflow Trainer ROADMAP 2025.

/cc @kubeflow/wg-training-leads @astefanutti @franciscojavierarceo @kramaranya @rudeigerc @mahdikhashan @Doris-xm @Harshal292004 @saileshd1402 @kannon92 @deepanker13

Love this feature?

Give it a 👍 We prioritize the features with most 👍

Guide contributeur