Installation¶

NeuroAtlas requires Python 3.10 or newer and runs on Linux.

Install the package¶

Install the package from PyPI into a new environment:

condavenv

conda create -n neuroatlas python=3.11 -y
conda activate neuroatlas
pip install neuroatlas-bench

python3.11 -m venv .venv
source .venv/bin/activate
pip install neuroatlas-bench

The install takes about 5 minutes and 6.5 GB of disk space. To install exactly the package versions used in the paper, add the constraints file:

pip install neuroatlas-bench -c https://raw.githubusercontent.com/kkontras/NeuroAtlas/v0.1.1/requirements-fm.txt

The distribution is called neuroatlas-bench; the command and the Python import are both neuroatlas.

Warning

Do not pip install neuroatlas. The package on PyPI called neuroatlas is unrelated to this project.

To work on the code, or to run run/default_runs.sh and the notebook in reproduction/, install from a clone instead:

git clone https://github.com/kkontras/NeuroAtlas.git
cd NeuroAtlas
pip install -e . -c requirements-fm.txt

Check the install:

neuroatlas --version

It prints neuroatlas 0.1.0. From a clone it also prints the commit.

GPU¶

The default PyTorch build needs a GPU with compute capability 7.5 or higher and a CUDA 13 driver. On older hardware, install a CUDA 12 build of PyTorch first. The probes themselves run on the CPU.

Extra packages¶

Some models and datasets need more packages:

To run	Install
EEG foundation models, supervised baselines, sleep and epilepsy datasets	the base install
Chronos	`pip install "neuroatlas-bench[ts]"`
MOMENT	the line above, then `pip install --no-deps "momentfm==0.1.4"`
Moirai	a separate environment from `requirements-tsfm.txt` (Python 3.10, torch 2.4.1)
the 14 MOABB BCI datasets	`pip install "neuroatlas-bench[bci]"`, then `pip install --no-deps "moabb==1.2.0"`

BCI datasets (MOABB)Chronos and MOMENTMoirai

pip install "neuroatlas-bench[bci]"
pip install --no-deps "moabb==1.2.0"

pip install "neuroatlas-bench[ts]"
pip install --no-deps "momentfm==0.1.4"

Moirai needs an older PyTorch and has its own environment (Python 3.10, torch 2.4.1), described in requirements-tsfm.txt.

--no-deps keeps the old pins that moabb 1.2.0 and momentfm declare from downgrading the rest of the stack. Both run correctly with it. pip check will still list those declared pins.

The base install includes xlrd and openpyxl, which read the ages of Sleep-EDF and ISRUC subjects for brain age. Dreyer2023 and Kim2025BetaRange are newer than moabb 1.2.0, so their readers ship inside the package.

Next¶

Quick start sets up a project folder and runs a first benchmark.