install.md

Installation

STREAMLINE can be run from Google Colab, a local notebook, or the command line. Local command-line and notebook use should be done from the repository root so Python can import the streamline package.

Google Colab

No local installation is required for Colab. Open the current notebook and run the setup cells:

Open the STREAMLINE Colab notebook

The notebook clones the repository with --depth 1, installs requirements, and exposes a parameter block for binary, multiclass, regression, and custom dataset runs.

Local Conda Environment

The recommended local setup is a dedicated conda environment:

STREAMLINE supports Python 3.10 and newer. Python 3.11 is the recommended default for local demos because it works well across the current scientific Python stack while remaining close to common notebook runtimes.

git clone --single-branch https://github.com/UrbsLab/STREAMLINE.git
cd STREAMLINE
conda create -n streamline python=3.11 pip
conda activate streamline
conda install pytorch=2.6 -y
pip install -r requirements.txt

Then confirm that the config runner is available:

python run.py --help

TabPFN is optional and is not required for the default STREAMLINE demos. Install it separately with pip install tabpfn before running TabPFN models. TabPFN also requires a Prior Labs token before model weights can be downloaded; see TabPFN Token Setup.

Local venv Environment

A standard Python virtual environment also works:

git clone --single-branch https://github.com/UrbsLab/STREAMLINE.git
cd STREAMLINE
python -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt

Building The Documentation

Install the docs-only packages and run Sphinx:

pip install -r docs/requirements.txt
sphinx-build -b html docs/source docs/build/html

The generated site opens from:

docs/build/html/index.html

Local Jupyter

Install Jupyter if it is not already available:

pip install jupyter
jupyter notebook

Open STREAMLINE_Notebook.ipynb from the repository root. The notebook keeps a parameter block at the top so the same notebook can run binary, multiclass, regression, or custom datasets.

Cluster Notes

Cluster use is environment-specific. The phase CLIs and config runner accept run_cluster settings such as Serial, Local, Parallel, BashSLURM, and BashLSF depending on phase support. For long runs, use a persistent terminal session such as tmux or screen so orchestration is not interrupted if your SSH connection drops.

Use the scheduler templates in run_configs/hpc/ as starting points:

python run.py -c run_configs/hpc/cedars_slurm_hcc.cfg --dry_run
python run.py -c run_configs/hpc/upenn_lsf_hcc.cfg --dry_run

See HPC and Cluster Runs for tmux basics, SLURM/LSF monitoring commands, scheduler config fields, and recovery notes.

Known Installation Issues

Some modeling and reporting packages include compiled dependencies. On macOS, especially on Apple Silicon or fresh Conda environments, install the compiled pieces through conda-forge first if pip reports build, linker, Graphviz, or WeasyPrint errors:

conda install -c conda-forge lightgbm xgboost catboost -y
conda install -c conda-forge graphviz python-graphviz -y
conda install -c conda-forge weasyprint cairocffi cairo pango gdk-pixbuf libffi -y
pip install -r requirements.txt

If the failure is isolated to one package, install only that package from conda-forge and rerun pip install -r requirements.txt.