Installing Spike Sorters

An important aspect of SpikeInterface is the spikeinterface.sorters module. This module wraps many popular spike sorting tools, allowing you to run multiple sorters on the same dataset with only a few lines of code and through Python.

Installing spike sorters can be painful! Many of them come with several requirements that could cause conflicts in your Python environment. To make things easier, we have created Docker images for most of these sorters, and in many cases the easiest way to run them is to do so via Docker or Singularity. This is the approach we recommend for all users. To run containerized sorters see our documentation here: Running sorters in Docker/Singularity Containers.

There are some cases where users will need to install the spike sorting algorithms in their own environment. If you are on a system where it is infeasible to run Docker or Singularity containers, or if you are actively developing the spike sorting software, you will likely need to install each spike sorter yourself.

Some of theses sorters are written in Matlab, so you will also need to install Matlab if you want to use them (Kilosort, Kilosort2, Ironclust, …). Some of then will also need some computing libraries like CUDA (Kilosort, Kilosort2, Ironclust (optional)) or opencl (Tridesclous) to use hardware acceleration (GPU).

Here is a list of the implemented wrappers and some instructions to install them on your local machine. Installation instructions are given for an Ubuntu platform. Please check the documentation of the different spike sorters to retrieve installation instructions for other operating systems. We use pip to install packages, but conda should also work in many cases.

Some novel spike sorting algorithms are implemented directly in SpikeInterface using the spikeinterface.sortingcomponents module. Checkout the SpikeInterface-based spike sorters section of this page for more information!

If you experience installation problems please directly contact the authors of these tools or write on the related mailing list, google group, GitHub issue page, etc.

Please feel free to enhance this document with more installation tips.

External sorters


  • Python + C++

  • Url:

  • Authors: Matthias Hennig, Jano Horvath,Cole Hurwitz, Oliver Muthmann, Albert Puente Encinas, Martino Sorbaro, Cesar Juarez Ramirez, Raimon Wintzer: GUI and visualisation

  • Installation:

    pip install herdingspikes


  • Matlab

  • Url:

  • Authors: Roland Diggelmann, Felix Franke

  • Installation:

    git clone
    # provide installation path by setting the HDSORT_PATH environment variable
    # or using HDSortSorter.set_hdsort_path()


  • Matlab

  • Url:

  • Authors: James J. Jun

  • Installation needs Matlab:

    git clone
    # provide installation path by setting the IRONCLUST_PATH environment variable
    # or using IronClustSorter.set_ironclust_path()






  • Python, requires CUDA

  • Url: /

  • Authors: Marius Pachitariu, Shashwat Sridhar, Alexander Morley, Cyrille Rossant, Kush Bunga

  • Install the python cuda toolkit. In principle, this should work:

    pip install cupy  (or pip install cupy-cudaXXX)
  • However, conda installation could be less painful:

    conda install cupy
  • Next, clone and install pykilosort. Note that we support the newer version on the develop branch and the ibl_prod version from the IBL fork:

    pip install phylib, pypandoc
    # recommended
    git clone --branch ibl_prod
    # or
    git clone --branch develop
    cd pykilosort
    pip install -r requirements.txt
    python install
  • Alternatively, you can use the pyks2.yml environment file in the pykilosort repo and update your favorite environment with:

    conda env update --name my-fav-env --file pyks2.yml --prune
  • See also




  • Python, requires MPICH

  • Url:

  • Authors: Pierre Yger, Olivier Marre

  • Installation:

    sudo apt install libmpich-dev
    pip install mpi4py
    pip install spyking-circus --no-binary=mpi4py


  • Python, runs faster with opencl installed but optional

  • Url:

  • Authors: Samuel Garcia, Christophe Pouzat

  • Installation:

    pip install tridesclous
  • Optional installation of opencl ICD and pyopencl for hardware acceleration:

    sudo apt-get install beignet (optional if Intel GPU)
    sudo apt-get install nvidia-opencl-XXX (optional if NVIDIA GPU)
    sudo apt-get install pocl-opencl-icd (optional for multi core CPU)
    sudo apt-get install opencl-headers ocl-icd-opencl-dev libclc-dev ocl-icd-libopencl1
    pip install pyopencl


  • Matlab

  • Also supports Snippets (waveform cutouts) objects (BaseSnippets)

  • Url:

  • Authors: Fernando Chaure, Hernan Rey and Rodrigo Quian Quiroga

  • Installation needs Matlab:

    git clone
    # provide installation path by setting the WAVECLUS_PATH environment variable
    # or using WaveClusSorter.set_waveclus_path()


  • Python

  • Url:

  • Authors: Johannes Niediek, Jan Boström, Christian E. Elger, Florian Mormann

  • Installation:

    git clone
    # Then inside that folder, run:
    # provide installation path by setting the COMBINATO_PATH environment variable
    # or using CombinatoSorter.set_combinato_path()

SpikeInterface-based spike sorters

Thanks to the spikeinterface.sortingcomponents module, some spike sorting algorithms can now be fully implemented with SpikeInterface.


This is a upgraded version of SpykingCircus, natively written in SpikeInterface. The main differences are located in the clustering (now using on-the-fly features and less prone to finding noise clusters), and in the template-matching procedure, which is now a fully orthogonal matching pursuit, working not only at peak times but at all times, recovering more spikes close to noise thresholds.

  • Python

  • Requires: HDBSCAN and Numba

  • Authors: Pierre Yger

  • Installation:

    pip install hdbscan
    pip install spikeinterface
    pip install numba  (or conda install numba as recommended by conda authors)


This is an upgraded version of Tridesclous, natively written in SpikeInterface. #Same add his notes.

  • Python

  • Requires: HDBSCAN and Numba

  • Authors: Samuel Garcia

  • Installation:

    pip install hdbscan
    pip install spikeinterface
    pip install numba

Legacy Sorters

Klusta (LEGACY)


  • Python, CUDA, torch

  • Requires SpikeInterface<0.96.0 (and Python 3.7)

  • Url:

  • Authors: JinHyung Lee, Catalin Mitelut, Liam Paninski

  • Installation: