This is a JAX-based Python toolbox for research and application of prototype-based machine learning methods and other interpretable models. All models are JIT-compiled and run on CPU, GPU and TPU. The focus of Prosemble is ease-of-use, extensibility and speed.
Prosemble can be installed using pip:
pip install prosembleTo install with JAX support:
pip install "prosemble[jax]" # CPU
pip install "prosemble[jax-cuda12]" # GPU (CUDA 12)To install the development version:
git clone https://github.com/naotoo1/prosemble.git
cd prosemble
pip install -e ".[all]"The full documentation is available at prosemble.readthedocs.io.
To build locally:
cd sphinx-docs && make htmlSee the examples/ directory.
Prosemble provides a reproducible development environment using devenv:
git clone https://github.com/naotoo1/prosemble.git
cd prosemble
devenv shell
uv sync --extra jax --extra dev
uv run pytest tests/ -x -q@misc{Otoo_Prosemble_2022,
author = {Otoo, Nana Abeka},
title = {Prosemble},
year = {2022},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/naotoo1/Prosemble}},
}