Installation#

All commands below use pip. For editable (development) installs from a local clone, pass the -e flag (e.g. pip install -e .).

Tip

We recommend uv as a fast, modern package manager. To use it, substitute uv pip for pip in all commands below. Note that uv pip install defaults to editable mode, so the -e flag is not needed.

PyPI#

Install the stable version of slmsuite from PyPI using:

pip install slmsuite

GitHub#

Install the latest version of slmsuite from GitHub using:

pip install git+https://github.com/holodyne/slmsuite

One can also clone slmsuite directly and add its directory to the Python path. Remember to install the dependencies (next sections).

git clone https://github.com/holodyne/slmsuite

Required Dependencies#

The following python packages are necessary to run slmsuite. These are listed as PyPI dependencies and thus are installed automatically if PyPI is used to install.

One can also install these dependencies directly by calling the following inside the package directory.

pip install -e .

Hardware Dependencies#

The following python packages are optional acceleration or hardware requirements, which the user can install selectively.

  • GPU pip install -e ".[gpu]"
    • cupy, highly recommended for GPU-accelerated holography. Sometimes, installation is made complicated by a pre-installed version of CUDA. You can find the CUDA version with nvcc --version in a terminal, and then install an installation of cupy specific to CUDA version YY with pip install cupy-cudaYYx.

  • Gradients pip install -e ".[torch]"
    • pytorch, required for conjugate gradient hologram optimization, either in GPU or CPU mode. Uses cupy - torch interoperability to pass data between modules without copying overhead, even on the GPU.

  • Cameras pip install -e ".[cameras]"
  • SLMs pip install -e ".[slms]"
    • pyglet

    • Other SLMs are loaded directly via .dll.

  • Image saving pip install -e ".[images]"
    • For most images and videos, imageio

    • Many video formats additionally require pyav

    • For .gif optimization, pygifsicle

Jupyter#

We highly recommended using Jupyter notebooks for interactive computing. Consider also using IPython magic, features like %autoreload 2 or %matplotlib inline.

If Jupyter is not used, the default matplotlib plots will block further execution, so the user should avoid plotting by using plot=False flags on functions.

Use the following to install recommended jupyter-related packages.

pip install -e ".[jupyter]"

All Dependencies#

To install all optional dependencies at once (including GPU, SLMs, cameras, images, Jupyter, docs, and testing), use the dev extra:

pip install -e ".[dev]"