Installation¶

warm-transfer has a lightweight core and optional extras for benchmark engines and neural methods. Python 3.11 or newer is required.

pipuv

pip install warm-transfer            # core
pip install "warm-transfer[bench]"   # benchmark donors and dataset tooling
pip install "warm-transfer[all]"     # bench + deep extras

uv add warm-transfer            # core
uv add "warm-transfer[bench]"   # benchmark donors and dataset tooling
uv add "warm-transfer[all]"     # bench + deep extras

The package is published on PyPI.

Variants¶

Variant	Install	Includes
Core	`warm-transfer`	`warmtransfer`: methods, metrics, similarity and public types
Benchmark	`warm-transfer[bench]`	donor engines, dataset downloads, parquet/YAML tooling and `warmbench`
Deep	`warm-transfer[deep]`	`torch` for neural cold-start methods such as `dropoutnet`
All	`warm-transfer[all]`	benchmark and deep dependencies

From source (development)¶

To work on warm-transfer itself, clone the repository and sync the environment:

uv sync                 # core + dev tools
uv sync --extra bench   # + benchmark donors and dataset tooling
uv sync --extra all     # + deep extras

Smoke checks¶

uv run python examples/quickstart.py
uv run warmbench --list-components

The first command checks the core package. The second checks that the benchmark entrypoint and registries are available.

Note

examples/quickstart.py is only present in a repository clone; it is not shipped in the pip wheel. For a pip install, use the inline snippet on the Quickstart page to smoke-check the core package instead.