Installation

You can install SToG from PyPI:

pip install stog

For development from source:

git clone https://github.com/intsystems/SToG.git
cd SToG
pip install -e .

Install Dependencies

For development with documentation:

pip install -r requirements.txt
pip install -r doc/requirements.txt

For experimentation and benchmarking:

pip install jupyter matplotlib seaborn pandas

Verify Installation

Test the installation:

import torch
from SToG import STGLayer, FeatureSelectionTrainer, create_classification_model

# Quick smoke test
model = create_classification_model(n_features=20, n_classes=2)
selector = STGLayer(n_features=20)

X = torch.randn(32, 20)  # batch_size=32, n_features=20
y = torch.randint(0, 2, (32,))

X_gated = selector(X)
print(f"Input shape: {X.shape}")
print(f"Gated output shape: {X_gated.shape}")
print("Installation successful!")

Or run tests:

python test/run_tests.py check
python test/run_tests.py all

Building Documentation

To build HTML documentation locally:

cd doc
make html

The documentation will be built in doc/build/html/. Open index.html in your browser.

Troubleshooting

Import Error: No module named ‘SToG’

Ensure you installed the package in development mode:

cd SToG
pip install -e .

PyTorch not found

Install PyTorch following the official guide at https://pytorch.org/get-started/locally/

For CPU-only installation:

pip install torch --index-url https://download.pytorch.org/whl/cpu

For CUDA support, refer to the official PyTorch documentation.

Documentation build fails

Reinstall documentation dependencies:

pip install --upgrade -r doc/requirements.txt