ML application for water potability classification

Dataset Info

• The source of the dataset is the following repo
• The task is a binary classification task to predict the water potability given the different feature measurements of the water quality
• The dataset sample contains about 3.2K samples

Project Info

• This repo contains a water potability Machine Learning FastAPI application deployment
• For the MLOps - MLFlow, FastAPI, Docker, Kubernetes has been utilized
• For deployment, an ML backend API app has been developed and deployed using FastAPI, docker, kubernetes
• For the training, the dataset is split into 90% - 10% for train and test sets respectively
• For getting the latest model from the mlflow logs for production, use the script src/get_model_for_production.py
• The python packages are listed in requirements.txt
• The docker container can be deployed using src/backend_app/Dockerfile
• For training and logging the model, use the src/modeling/ml_model_dev.py script
• The FastAPI app deployment code is in src/backend_app/app.py script
• The Kubernetes deployment configuration files can be found in src/deploy/
• To test the deployed FastAPI app on a local machine, the src/test_post_request.py script can be used
• This project involved learning orchestrating the scaling of the deployed ML service app with kubernetes

Docker deployment of ML backend app on a local machine

• To build the container for the backend app, run the following command from the repo root

docker build -f src/backend_app/Dockerfile -t ml-water-potability-backend .

• Use --no-cache in the above command to build the image without using cached layers
• To the run the container with the backend app, run the following command

docker run -p 5000:5000 -t ml-water-potability-backend

Kubernetes deployment of ML backend app on a local machine

• The deployment architecture used in this project is Serverful (Raw deployment)
• To deploy the container on a kubernetes cluster, refer src/deploy/README.md for detailed instructions
• Alternatively, run src/deploy/start_deployment.sh for starting the kubernetes deployment

HuggingFace deployment

• The FastAPI application with appropriate changes has also been deployed to HuggingFace
• To test the deployed FastAPI app on HuggingFace, use the src/test_post_request.py script in the HuggingFace repo since the endpoint is different

Documentation

The documentation generated with sphinx is available in docs/_build/html/index.html