Skip to content

README.md

Latest commit

 

History

History
56 lines (45 loc) · 2.02 KB

README.md

File metadata and controls

56 lines (45 loc) · 2.02 KB

SkyPilot Integration with Nebius AI Cloud

Overview

SkyPilot is an open-source framework for running AI and batch workloads. Nebius AI Cloud offers seamless integration with SkyPilot, simplifying the process of launching and managing distributed AI workloads on powerful GPU instances:

https://nebius.com/blog/posts/nebius-ai-cloud-skypilot-integration

Prerequisites

Before getting started, ensure you have:

  • Nebius Account and CLI:

    • Create your Nebius account
    • Install and configure the Nebius CLI
    • Create a standalone Skypilot instance on Nebius.
    • Prepare a configuration file location in ~/.sky/config.yaml. Skypilot Config Documentation 
      nebius:
  region_configs:
    <your_region>:
      project_id: project-<your_project_id>
      filesystems:
      - filesystem_id: computefilesystem-<your_filesystem_id>
        mount_path: <your_mount_path>
        attach_mode: READ_WRITE
api_server:
  endpoint: https://USER:PASSWORD@<skypilot_instance_public_ip_endpoint>
    • Download and run the setup script: 
      wget https://raw.githubusercontent.com/nebius/nebius-solution-library/refs/heads/main/skypilot/nebius-setup.sh
chmod +x nebius-setup.sh 
./nebius-setup.sh
      • You'll be prompted to select a Nebius tenant and project ID

  • Python Requirements:

    • Python version 3.10 or higher
    • Install SkyPilot with Nebius support: 
      pip install "skypilot-nightly[nebius]"

Examples and Solutions

For detailed examples and solutions using SkyPilot on Nebius, refer to https://github.com/nebius/ml-cookbook/tree/main/skypilot

It includes examples on how to:

  • lanch individual VMs as well as clusters
  • mount Nebius Object Storage to SkyPilot clusters
  • do distibuted training and inference
  • etc.

For detailed SkyPilot documentation, refer to the official SkyPilot documentation.