ML Production Scaffold

This repository provides a scaffold for deploying Machine Learning models into production, including project structure, APIs, and deployment pipelines.

Architecture Component Breakdown

The proposal is a "High-Level Architecture" diagram for a machine learning system, consisting of four layers:

Evaluation Layer - This layer ensures the equivalence of two models and monitors production models by comparing live traffic predictions with training predictions. Scoring Layer - Converts features into predictions. Standard libraries for this layer include Scikit-learn, XGBoost, and Keras. Feature Layer - Responsible for generating reusable, transparent, and scalable feature data. Data Layer - Provides access to all data sources, simplifying data reproducibility challenges.

Each layer is visually represented with green boxes and arrows pointing to descriptions. The diagram is sourced from Building a Reproducible Machine Learning Pipeline by Sugimura and Hartl.

Workflow Summary

Clients send requests to the REST API for predictions. The API utilizes the Feature Extractor and the Trained Model to generate predictions. The CI/CD Pipeline automates updates, ensuring new models and feature extraction logic remain consistent between training and production. This architecture balances real-time inference with a robust development workflow for continuous model improvement.

Repository Structure

ml_production_scaffold/
│-- regression_model/
│   │-- config/   # General configurations
│   │-- datasets/ # Model input data
│   │-- processing/  # Model definition
│   │-- trained_models/  # Trained Models
│   │-- config.yml  # General configurations
│   │-- pipeline.py  # Preprocessing pipeline
│   │-- predict.py  # Inference code
│   │-- train_pipeline.py  # Training script
│   │-- VERSION
│-- api/
│   │-- app/
│   │   │-- tests/
│   │   │-- schems/
│   │   │-- main.py
│   │   │-- config.py
│   │   │-- api.py
│-- requirements/
│   │-- requirements.txt  # Requirements file
│   │-- test_requirements.txt  # Requirements to test environment
│-- tests/ # Unit tests
│-- setup.py  # setup project
│-- MANIFEST.in
│-- pyproject.toml  # Project dependencies
│-- tox.ini  # Project dependencies
│-- README.md  # Project documentation

Requirements

Ensure you have Python 3.8+ installed and run:

pip install -r requirements.txt

Model Training

To train the model:

python src/train.py

Running the API

To start the API using FastAPI:

uvicorn api.main:app --host 0.0.0.0 --port 8000

Testing

To run the tests:

pytest tests/

Deployment with Docker

To build and run the container:

docker build -t ml_model .
docker run -p 8000:8000 ml_model