An ecosystem for biological foundation models.
Create and activate a virtual environment:
uv venv
source .venv/bin/activatePre-requisites:
- Pytorch (e.g.
uv pip install torch)
Install the package:
uv pip install -e .# PlantCaduceus
uv pip install -e .[mamba]
# GPN
uv pip install -e .[gpn]
# GLM-Experiments
uv pip install -e .[glm-experiments]
# Then manually install glm-experiments to avoid circular dependency:
uv pip install git+https://github.com/Open-Athena/glm-experimentsFor evo2 support, Docker is recommended as it handles all complex CUDA/cuDNN dependencies automatically.
See docker/evo2/README.md for detailed instructions.
Quick start:
cd docker/evo2
./run.sh local # or ./run.sh githubRun the example scripts:
source .venv/bin/activate
# light job
python examples/marin_evolutionary_constraint.py
# heavier job, split across 8 GPUs with 8 CPU threads per GPU
OMP_NUM_THREADS=8 torchrun --nproc_per_node=8 examples/plantcad_llr.pyTo set up the development environment with linting, formatting, type checking, and testing:
# Install development dependencies
uv pip install --group dev
# Install pre-commit hooks
pre-commit install# Run all pre-commit hooks (linting, formatting, type checking)
pre-commit run# Run all tests
pytest