Customised Workstations

Containerization of the Third Party Algorithm

Objective

Containerizing third party algorithms and using Zeblok's Ai-MicroCloud™ to provide access to more audiences.

Steps to follow

Zeblok offers a basic minimal notebook with pre-installed CUDA Libraries and the associated Docker Hub URL and Dockerfile Urls are given below.

https://hub.docker.com/u/zeblok (Docker hub) https://github.com/zeblok/dockerImagesFiles (Dockerfiles on github)

• Developers can use one of the Zeblok’s ready to use Docker Images as a base Image in his/her Dockerfile.

• Developers can write the instructions to install dependencies that are required by the Algorithm. (For Instance, if a user’s Algorithm requires BlazingSQL and Rapids AI as dependency, the user can take the Minimal notebook as base and install the dependencies on top of that base in a new Dockerfile as shown below).

ARG BASE_CONTAINER=zeblok/minimal-notebook:v0.0.7
FROM $BASE_CONTAINER
USER $NB_UID
RUN pip install -q --pre bentoml
RUN pip install 'scikit-learn>=0.23.2' 'pandas>=1.1.1'
USER $NB_UID

• After writing the Instructions for installing the dependencies, the developer needs to build the Image from the Dockerfile using the command `docker build -t ‘ALGO_NAME:VERSION’ .

• After Building the image user can push the image to dockerHub using the command

  `docker push dockerUsername/ALGO_NAME:VERSION`.

• After successfully pushing the Docker Image developer will share the docker image Tag, short description, and an Image (400x220px) to [e-mail].