Skip to content

Latest commit

 

History

History
38 lines (31 loc) · 1.53 KB

README.md

File metadata and controls

38 lines (31 loc) · 1.53 KB
Raw

KAN AI Container

(last edit: 4/24/2023)

KAN AI container is a Docker container that gets deployed when you deploy your AI solution to a computer device. It runs as an IoT Edge module when deployed to Azure IoT Edge, or as a Kubernetes pod when deployed to Kubernetes. You can use the provided devops.py Python script to build this container image yourself.

0. Prerequisites

  • Docker

    NOTE: Some tools we use work better with access to docker commands without sudo. Use Docker Desktop version when possible. Otherwise, you need to add your user to docker group (see instructions). Please don't use rootless model, which isn't supported by some of the tools.

  • Python

1. Define environment variables

Define an environment variable that points to your container registry name:

export CONTAINER_REGISTRY_NAME=kanprod.azurecr.io # replace with your own container registry name

2. Install required packages

pip install -r requirements.txt

3. Build all the images

python devops.py build-all

The above command build four KAN AI container images for different infrastructures:

Image name Infrastructure
kanai:<version>-openvionamd64 x64 CPU with OpenVino
kanai:<version>-gpuamd64 x64 CPU with Nvidia GPU
kanai:<version>-amd64 x64 CPU only
kanai:<version>-jetson Nvidia Jetson platform

4. Push images

python devops.py push-all