WG Training focuses on the development, deployment, and operation of distributed AI workloads on Kubernetes. It covers the entire model development lifecycle, including pre-training, post-training (such as LLM fine-tuning), hyperparameter optimization, reinforcement learning, and other techniques required to build, train, and run AI models at scale.

The charter defines the scope and governance of the Training Working Group.

• Kubeflow Trainer and Katib Call: Wednesdays at 08:00AM PT (Pacific Time) (bi-weekly). Convert to your timezone.
  • Meeting notes and Agenda.
  • Meeting recordings.

• Andrey Velichkevich (@andreyvelich), Apple
• Antonin Stefanutti (@astefanutti), Red Hat
• Yuki Iwai (@tenzen-y), Preferred Networks, Inc.
• Yuan Tang (@terrytangyuan), Red Hat

• Slack: #kubeflow-trainer
• Mailing list
• GitHub Teams:
  • @kubeflow/wg-training-chairs - Team of Training Working Group Chairs

The following subprojects are owned by WG Training:

• Owners:
  • https://raw.githubusercontent.com/kubeflow/katib/master/OWNERS

• Owners:
  • https://raw.githubusercontent.com/kubeflow/mpi-operator/master/OWNERS

• Owners:
  • https://raw.githubusercontent.com/kubeflow/trainer/master/OWNERS
• Contact:
  • GitHub Teams:
    • @kubeflow/kubeflow-trainer-team - Kubeflow Trainer maintainers