Installation

Library and CLI

Python 3.11 or later is required. For a released package from PyPI:

python -m pip install lrdbench

For the current development source, clone the repository and install from its root:

pip install -e .

Development tooling:

pip install -e ".[test,dev]"

Documentation build (local)

Install documentation dependencies and serve the site:

pip install -e ".[docs]"
mkdocs serve

The site configuration is mkdocs.yml at the repository root. Read the Docs builds it using .readthedocs.yaml at lrdbench.readthedocs.io. GitHub Pages builds it with mkdocs.pages.yml at dave2k77.github.io/lrdbench.

Optional extras

Extra	Purpose
reports	Jinja2 / tabulate helpers for richer reporting
parquet	PyArrow dependency for downstream use; the public result store currently writes CSV
ml	scikit-learn baselines: MLRandomForest, MLSVR
nn	PyTorch baselines: MLCNN, MLLSTM
data-driven	All ML and NN baseline dependencies
docs	MkDocs + Material + mkdocstrings
test	pytest, coverage, Hypothesis
dev	Ruff, mypy, build, twine, pre-commit
all	Reports, parquet, docs, notebooks, tests, dev tools, and all data-driven dependencies

Example: pip install -e ".[docs,test]".

For a full local development environment, including the heavier scikit-learn and PyTorch data-driven dependencies, use:

pip install -e ".[all]"

For a lighter development environment that avoids PyTorch, prefer .[test,dev,docs,reports].

Because the repository's default pytest configuration enables coverage reports, install the test extra before running python -m pytest. For a quick no-coverage smoke check in a minimal environment, run python -m pytest -q -o addopts=''.

Data-driven smoke benchmarks need the feature-based ML extra and reporting support:

pip install -e ".[ml,reports]"
lrdbench run configs/suites/smoke_data_driven.yaml

The neural-network extra may install a large PyTorch distribution depending on your platform. Use ml for the packaged RF/SVR smoke suite, and install nn only when running CNN/LSTM benchmarks.