Effective-MLOPS for Machine learning

Overview

This is an implementation of CI/CD in an ML project using GItOps and Experiment tracking. The focus is on utilizing GitHub Actions to enforce and automate certain policies and Weights & Bias to generate reports on model runs.

Motivation

This project is the result of me taking the weights and bias course on Ci/CD for MLOPs and reading and digging more about MLOPs and realizing I'll like to help others have a resource which they can refer too when trying to learn about. You can find a complete blog post about this on Medium

File structure

.github/workflows: contains all GitHub Actions workflows files. /data: Train and testing data /notebooks: Notebooks for model development and expirement. /source/tests: python scripts to geenrate reports, tests etc requiremnts.txt

How to install and run project

• Fork this repository
• Clone your forked repository
• create Wandb account get API keys
• Use Wand credentials in the notebook and run notebook
• Create secret for Wandb Api key
• Commit and push with comment '/wandb

How to use the project to setup your CI/CD workflow

• Copy the workflow in .github/workflows folder and edit accord to your desired workflow
• Copy model_report.py to generate report