sk-serve

Deployment of a Scikit-Learn model and it's column transformations with a single endpoint. Only a traditional Scikit-Learn model is needed and a ColumnTransformer object (sklearn.compose) to deploy your model. Validation of input data is also supported with pydantic.

Usage

See the Examples section of the repository.

Installation

The package exists on PyPI (with a different name though) so you can install it directly to your environment by running the command

pip install simple-serve

Dependencies

• pydantic
• fastapi
• pandas
• scikit-learn

Additional packages for development:

• pyright
• pre-commit

Development

If you want to contribute you fork the repository and clone it on your machine

git clone https://github.com/alexliap/sk_serve.git

And after you create you environment (either venv or conda) and activate it then run this command

pip install -e ".[dev]"

That way not only the required dependencies are installed but also the development ones.

Also this makes it so that when you import the code to test it, you can do it like any other module but containing the changes you made locally.

Before you decide to commit, run the following command to reformat code in order to be in the acceptable style.

pre-commit install
pre-commit run --all-files