Deployment example for a scikit-learn/lightgbm pipeline

This project shows a simple machine learning pipeline that uses scikit-learn and lightgbm. We train a model based on the New York Taxi trip dataset and then deploy it using FastAPI.

Development

This package is intended to be developed inside a conda environment defined with the environment.yml. With the following commands you can setup the development environment initially:

mamba env create
conda activate quantcore-reproduce
python -m pip install --no-deps --disable-pip-version-check -e .

Deployment

As this repository is an example for various ways to deploy a pipeline inside a docker container, there are several flavour present. We have supplied a small Makefile that can be used to execute the docker build command for each of them.

# Build the container
make anaconda
# Run the container
docker run -t nyc-taxi-anaconda -p 8000:8000