VirtML is a project dedicated to simplifying the process of setting up Kubeflow in a virtualized environment, making it easier for data scientists to leverage the power of Machine Learning (ML) workflows on Kubernetes.

Follow the guides below to get started with VirtML! You have two options:

👨‍🔬 Manual Deployment: Follow the step-by-step guides to set up a virtualized environment for Kubeflow using a manual approach, setting everything by hand. This is a great way to learn the ins and outs of the process.

🤖 Automated Deployment: Use Ansible playbooks to automate the deployment of a virtualized environment for Kubeflow. This is a great way to quickly get started with VirtML. The automated deployment is still a work in progress, but it will be available soon.

In any case, you should complete the core guides first, as they lay the foundation for either approach.

Note

This project is tested on Debian 12 (Bookworm). It should work on other Debian-based distributions but it will require some modifications. We also plan to support Rocky Linux in the future.

🚧 🚧 🚧 This is Work-in-Progress 🚧 🚧 🚧

Core Guides

• Install QEMU and libvirt on Debian-based Systems
• Create e VirtML Administrative VM

Deployment Guides

• Manual Deployment
  • Configure a PXE server
  • Create a Kubernetes Node
  • Deploy Kubernetes
  • Deploy Longhorn
  • Deploy Kubeflow

System Guides

• GPU Passthrough using VFIO

Contact

• Dimitris Poulopoulos dimitris.a.poulopoulos@gmail.com

We warmly welcome your feedback and look forward to hearing from you!