#Introduction

#What is HAMi?

Heterogeneous AI Computing Virtualization Middleware (HAMi), formerly known as k8s-vGPU-scheduler, is an "all-in-one" chart designed to manage heterogeneous AI computing devices in a Kubernetes cluster. It enables heterogeneous AI devices to be shared among tasks.

#Why HAMi

#Device sharing

• Supports multiple heterogeneous AI computing devices
• Supports device-sharing for multi-device containers

#Device Memory Control

• Hard limit inside the container
• Supports dynamic device memory allocation
• Supports memory allocation by MB or by percentage

#Device Specification

• Supports specifying a device type for heterogeneous AI computing devices
• Supports selecting a specific heterogeneous AI computing device by UUID

#Easy to try

• Transparent to tasks inside the container
• Install/uninstall via Helm — easy and clean

#Open and Neutral

• Jointly initiated by contributors from Internet, finance, manufacturing, and cloud providers
• Targets open governance through CNCF