autora-theorist-darts 1.0.0a1

Differentiable Architecture Search theorist for AutoRA

AutoRA DARTS Theorist

autora-theorist-darts is a Python module for fitting data using differentiable architecture search, built on AutoRA.

Website: https://autoresearch.github.io/autora/

User Guide

You will need:

• python 3.8 or greater: https://www.python.org/downloads/
• graphviz (optional, required for computation graph visualizations): https://graphviz.org/download/

Install DARTS as part of the autora package:

pip install -U "autora[theorist-darts]"

It is recommended to use a python environment manager like virtualenv.

Check your installation by running:

python -c "from autora.theorist.darts import DARTSRegressor; DARTSRegressor()"

Developer Guide

Get started

Clone the repository (e.g. using GitHub desktop, or the gh command line tool) and install it in "editable" mode in an isolated python environment, (e.g. with virtualenv) as follows:

In the repository root, create a new virtual environment:

virtualenv venv

Activate it:

source venv/bin/activate

Use pip install to install the current project (".") in editable mode (-e) with dev-dependencies ([dev]):

pip install -e ".[dev]"

Run the test cases:

pytest tests/ --doctest-modules src/

Activate the pre-commit hooks:

pre-commit install

Add new dependencies

In pyproject.toml add the new dependencies under dependencies

Install the added dependencies

pip install -e ".[dev]"

Publish the package

Update the metadata under project in the pyproject.toml file to include name, description, author-name, author-email and version

• Follow the guide here: https://packaging.python.org/en/latest/tutorials/packaging-projects/

Build the package using:

python -m build

Publish the package to PyPI using twine:

twine upload dist/*