Installing Fast TICC

If you’re reading this message, congratulations, you’re an early adopter! The code is still hot from being uploaded and we’re working on the PyPI and conda-forge recipes in real time.

From PyPI

Once this code is officially released, we’ll upload wheels to PyPI so that you can install with pip install fast_ticc.

From conda-forge

We’re going to contribute a recipe to conda-forge so that you can install with conda install -c conda-forge fast_ticc.

From source

First, get a copy of the source. Our Github repository is at https://github.com/sandialabs/fast_ticc and you can either clone or download the repository or download the release package for the latest version.

Second, if you haven’t done so already, create a Python virtual environment to hold the fast_ticc installation and its dependencies. This helps avoid version conflicts between different packages.

Third, go to the directory containing the package (whether you’ve cloned it or unpacked it from a download) and run python -m pip install .. This will build and install Fast TICC and its dependencies.

Developer Mode: If you want to work on the TICC code to add features or fix bugs, you can install it with python -m pip install -e .. The “-e” argument tells Python to use the source tree you downloaded instead of copying the library into its package collection.

PyPy, CPython, and Fast TICC

We support PyPy as well as the CPython interpreter. We depend on scikit-learn and NumPy for most of our math operations. Other relatively recent Python interpreters (version 3.8 or newer) will probably work as long as they support those two packages.

If you’re using CPython, Fast TICC will try to use Numba to accelerate some of its operations. If Numba isn’t available for whatever reason the code will still work fine.