How To Build and Deploy a Machine Learning Model with FastAPI

Abstraction: Wrapping and serving scikit-learn models as REST APIs with FastAPI

Key points:

• FastAPI is a modern Python 3.6+ web framework comparable in performance to NodeJS and Go; run with Uvicorn ASGI server via pip install fastapi uvicorn
• Built-in interactive Swagger UI documentation at /docs and ReDoc at /redoc — endpoints can be tested directly in the browser
• ML model pattern: train a Random Forest on Iris dataset, persist with joblib, wrap predictions in a Pydantic BaseModel class, expose via POST endpoint
• POST endpoints are preferred over GET for ML APIs — parameters sent as JSON body rather than URL query strings
• The same deployment pattern applies equally to simple sklearn models and complex neural networks
• Automatic API reload on file save with --reload flag; supports async programming for more advanced use cases

Connections: Fastapi · Model Deployment · Rest API · Machine Learning

Source: https://towardsdatascience.com/how-to-build-and-deploy-a-machine-learning-model-with-fastapi-64c505213857