Installation

Leonardo requires Python>=3.10. We highly recommend using conda virtual environment to install and operate Leonardo.

PyPI

CPU version

To use Leonardo with CPU, install Leonardo using:

conda create -n leonardo python=3.10
conda activate leonardo
pip install leonardo_toolset

Additional Napari plugins can be installed separately:

pip install lsfm_destripe_napari
pip install lsfm_fusion_napari

Leonardo has now been tested on Linux and Windows. Leonardo may have issues on macOS caused by third-party dependencies, specifically resulting in a “metadata-generation-failed” error.

Note

For Leonardo-DeStripe, if you are using the GPU-enabled version of JAX as backend on a machine without CUDA or GPU support, JAX may fail to initialize and raise an error when imported.

To avoid this, set the following environment variable before running your code to force JAX to use CPU only:

Linux/macOS

export JAX_PLATFORMS=cpu

Windows (CMD)

set JAX_PLATFORMS=cpu

For more context, see the official JAX discussion.

GPU version

To use Leonardo with GPU:

• Setup Pytorch according to your own system setting, following the official guideline.

• Setup Jax according to your own system setting, following the official guideline (Optional).

• Install Leonardo following the instructions under Section CPU version.

Development Version

To work with the latest development version, install from GitHub using:

pip install git+https://github.com/peng-lab/leonardo_toolset.git