Common Install Notes

A basic requirement to use Brainstorm is Numpy, and we recommend that you make sure that you have a fast BLAS installation which Numpy will use. OpenBLAS is excellent, and can be installed on a Debian family system easily: sudo apt-get install libopenblas-dev.

Brainstorm provide a PyCudaHandler which can be used to accelerate neural networks using Nvidia GPUs. In order to use it, you need to have CUDA 7.0 or later already installed and setup from

Installation variants

When installing from PyPI or GitHub, you can specify the following installation variants to additionally install optional dependencies:

all pycuda tests live_viz draw_net


Install prerequisites:

sudo apt-get install python-dev libhdf5-dev libopenblas-dev

Install the latest stable release from PyPI, including all optional dependencies:

pip install brainstorm[all]

which will additionally install pycuda, scikit-cuda, pygraphviz and bokeh.

To install the latest master branch, you can do:

pip install[all]


Instructions coming soon.


Instructions coming soon.


Builds of Brainstorm are available as Docker images: Brainstorm Docker (CPU) or Brainstorm Docker (CUDA). These are updated on a weekly basis with bleeding-edge builds of Brainstorm. Examples of running bash in a Docker container are as follows:

sudo docker run -it kaixhin/brainstorm
sudo nvidia-docker run -it kaixhin/cuda-brainstorm:7.0

For a guide to Docker, see the official docs. CUDA support requires NVIDIA Docker. For more details on how to use the Brainstorm Docker images, consult the source project.