Installation¶
Requirements¶
- Python ≥ 3.13
- PyTorch ≥ 2.2
- Linux / macOS / Windows
- Optional CUDA-enabled GPU for training speed
Install from PyPI¶
pip install neuroscience-control
Or with uv:
uv add neuroscience-control
Install from source¶
git clone https://github.com/dani2442/neuroscience_control.git
cd neuroscience_control
uv sync
Optional dataset integrations¶
nilearn is included by default. For OpenNeuro and DataLad dataset download support, install the datasets extra:
pip install "neuroscience-control[datasets]"
# or
uv sync --group datasets
Verify installation¶
python -c "import src; print(src.__version__)"
Import namespace¶
Install the package as neuroscience-control, but import from src in the current repository layout:
from src.models import CoupledHopfModel, NeuralSDE
from src.metrics import FCCorrelation, PhFCD
from src.training import Trainer, HopfConfig
Dependency notes¶
Complex-valued torchsde fork
This project depends on a fork of torchsde
that adds complex Brownian motion support. When installing from source with uv, this
is resolved automatically via the [tool.uv.sources] override in pyproject.toml.
If you install with plain pip, it will use the PyPI torchsde release (which may
lack complex-valued support).