Installation

Installation#

BrainPy is designed to run cross platforms, including Windows, Linux, and MacOS. It only relies on Python libraries.

Installation with pip #

You can install BrainPy from the pypi. To do so, use:

pip install brainpy

To update the latest BrainPy, you can use

pip install -U brainpy

If you want to install the pre-release version (the latest development version) of BrainPy, you can use:

pip install --pre brainpy

Installation from source #

If you decide not to use pip, you can install BrainPy from GitHub, or OpenI.

To do so, use:

pip install git+https://github.com/PKU-NIP-Lab/BrainPy

# or

pip install git+https://git.openi.org.cn/OpenI/BrainPy

Dependency 1: NumPy #

In order to make BrainPy work normally, users should install several dependent Python packages.

The basic function of BrainPy only relies on NumPy, which is very easy to install through pip or conda:

pip install numpy

# or

conda install numpy

BrainPy relies on JAX. JAX is a high-performance JIT compiler which enables users to run Python code on CPU, GPU, and TPU devices. Core functionalities of BrainPy (>=2.0.0) have been migrated to the JAX backend.

Linux #

Currently, JAX supports Linux (Ubuntu 16.04 or later) and macOS (10.12 or later) platforms. The provided binary releases of jax and jaxlib for Linux and macOS systems are available at

If you want to install a CPU-only version of jax and jaxlib, you can run

pip install --upgrade "jax[cpu]"

If you want to install JAX with both CPU and NVidia GPU support, you must first install CUDA and CuDNN, if they have not already been installed. Next, run

# CUDA 12 installation
# Note: wheels only available on linux.
pip install --upgrade "jax[cuda12_pip]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

# CUDA 11 installation
# Note: wheels only available on linux.
pip install --upgrade "jax[cuda11_pip]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

Alternatively, you can download the preferred release “.whl” file for jaxlib from the above release links, and install it via pip:

pip install xxx-0.4.15-xxx.whl

pip install jax==0.4.15

Note

Note that the versions of jaxlib and jax should be consistent.

For example, if you are using jax==0.4.15, you would better install jax==0.4.15.

MacOS #

If you are using macOS Intel, we recommend you first to install the Miniconda Intel installer:

Download the package in the link https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-x86_64.pkg
Then click the downloaded package and install it.

If you are using the latest M1 macOS version, you’d better to install the Miniconda M1 installer:

Download the package in the link https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-arm64.pkg
Then click the downloaded package and install it.

Finally, you can install jax and jaxlib as the same as the Linux platform.

pip install --upgrade "jax[cpu]"

Windows #

For Windows users with Python >= 3.9, jax and jaxlib can be installed directly from the PyPi channel.

pip install jax jaxlib

For Windows users with Python <= 3.8, jax and jaxlib can be installed from the community supports. Specifically, you can install jax and jaxlib through:

pip install "jax[cpu]" -f https://whls.blob.core.windows.net/unstable/index.html

If you are using GPU, you can install GPU-versioned wheels through:

pip install "jax[cuda111]" -f https://whls.blob.core.windows.net/unstable/index.html

Alternatively, you can manually install you favourite version of jax and jaxlib by downloading binary releases of JAX for Windows from https://whls.blob.core.windows.net/unstable/index.html . Then install it via pip:

pip install xxx-0.4.15-xxx.whl

pip install jax==0.4.15

WSL #

Moreover, for Windows 10+ system, we recommend using Windows Subsystem for Linux (WSL). The installation guide can be found in WSL Installation Guide for Windows 10/11. Then, you can install JAX in WSL just like the installation step in Linux/MacOs.

Dependency 3: brainpylib #

Many customized operators in BrainPy are implemented in brainpylib. brainpylib can also be installed from pypi according to your devices. For windows, Linux and MacOS users, brainpylib supports CPU operators. You can install CPU-version brainpylib by:

# CPU installation
pip install --upgrade brainpylib

For Nvidia GPU users, brainpylib only support Linux system and WSL2 subsystem. You can install the CUDA-version by using:

# CUDA 12 installation
pip install --upgrade brainpylib-cu12x

# CUDA 11 installation
pip install --upgrade brainpylib-cu11x

Running BrainPy with docker #

If you want to use BrainPy in docker, you can use the following command to pull the docker image:

docker pull brainpy/brainpy:latest

You can then run the docker image by:

docker run -it --platform linux/amd64 brainpy/brainpy:latest

Please notice that BrainPy docker image is based on the ubuntu22.04 image, so it only support CPU version of BrainPy.

Running BrainPy online with binder #

Click on the following link to launch the Binder environment with the BrainPy repository:

Wait for the Binder environment to build. This might take a few moments.

Once the environment is ready, you’ll be redirected to a Jupyter notebook interface within your web browser.