Help make zamba better

zamba is an open source project, which means you can help make it better!

Develop the GitHub repository

To get involved, check out the GitHub code repository. There you can find open issues with comments and links to help you along.

zamba uses continuous integration and test-driven development to ensure that we always have a working project. So what are you waiting for? git going!

Installation for development

To install zamba for development, clone the git repository and install the package with the developer dependency group. We recommend uv for managing the environment.

With uv (recommended):

$ git clone https://github.com/drivendataorg/zamba.git
$ cd zamba
$ uv pip install -e ".[image,video]" --group dev

When installing from a clone, [tool.uv] in pyproject.toml already sets torch-backend = "auto" and the protobuf / setuptools overrides needed for both image and video extras. End users installing from PyPI need to pass --torch-backend=auto and --overrides manually; see Installing zamba.

Or use the Makefile target (which uses uv):

$ make requirements

With pip:

$ pip install -e ".[tests,image,video,docs]"

If your contribution is to the DensePose model, install the DensePose dependencies from GitHub as described in the DensePose installation section.

To build the documentation locally with uv, install the docs dependency group (e.g. uv pip install -e . --group docs). With pip, install the docs extra (e.g. pip install -e ".[docs]").

Running the zamba test suite

The included Makefile contains code that uses pytest to run all tests in zamba/tests.

The command is (from the project root):

$ make test

For DensePose related tests, install the DensePose dependencies from GitHub (see DensePose installation), then run:

$ make test-densepose

Submit additional training videos

If you have additional labeled videos that may be useful for improving the basic models that ship with zamba, we'd love to hear from you! You can get in touch at info@drivendata.org