Installation#

spora [bench] runs from a clone of its repository. It depends on the data loading interface spora [io], which you install first.

1. Install spora [io]#

git clone https://github.com/bunnelab/spora-io.git
cd spora-io
pip install -e .

See the spora [io] installation guide for the SPORA_DATASETS_DIR setting and version requirements.

2. Install spora [bench]#

git clone https://github.com/bunnelab/spora-bench.git
cd spora-bench
pip install -r requirements.txt

The requirements include PyTorch, scikit-learn, Hydra/OmegaConf, AnnData, and torch-scatter. The cell phenotyping pipeline uses cuML’s logistic regression on a compatible GPU when it is installed and falls back to scikit-learn otherwise.

3. Set Up the Models#

Make sure the foundation models you want to evaluate are installed. For instructions, refer to their repositories:

  1. VirTues: https://github.com/bunnelab/virtues

  2. KRONOS: https://github.com/mahmoodlab/KRONOS

Model wrappers for the other entries of the Model Zoo import their own packages; install those for the models you intend to run. spora [bench] can also evaluate new models; see Add New Models.

All pipelines read configs/base_config.yaml relative to the working directory, so run them from the repository root.