mlops

Cloud agnostic solution for deploying Machine Learning models in docker and kubernetes

Dockerfile

This file is used to build Docker Image where our inferencing logic and trained model is present

predit.py

A simple python code that loads the trained model for the predictions based on the request

app.py

REST API logic using Flask where our predict.py code is exposed as an API

config.json

• registry-name : Name of the model registry.
• model-version: Version of the model registered.
• ecr-version: Docker Image version. This is used to deploy a model from Git Actions run use default in case if wanted to deploy a latest version.
• ram: Memory needed to provide values in MB
• CPU: CPU value provides the value from 0.1
• deployment-type: use the keyword deploy to deploy the model after successful completion of Git Actions or use Build to build only docker Image
• model_name: name of the trained ML model