Installation¶

pyTomoAO requires Python 3.9 or newer and runs on Linux, macOS and Windows. It is tested on 3.9 through 3.13.

From PyPI¶

pip install pyTomoAO

From source¶

git clone https://github.com/KeckObservatory/pyTomoAO.git
cd pyTomoAO
pip install .

Use an editable install if you intend to modify the code:

pip install -e .

In a fresh environment¶

Working in an isolated environment avoids clashes with other scientific stacks:

venv

python -m venv .venv
source .venv/bin/activate      # Windows: .venv\Scripts\activate
pip install pyTomoAO

conda

conda create -n pytomoao python=3.11
conda activate pytomoao
pip install pyTomoAO

Dependencies¶

The following packages are installed automatically:

Package	Used for
`numpy`	Array handling and linear algebra throughout
`scipy`	Sparse gradient matrices and the Cholesky solve
`numba`	JIT-compiled covariance kernels on the CPU path
`PyYAML`	Reading the configuration file

matplotlib is not installed automatically. It is needed only by visualize_reconstruction, visualize_commands and the display=True branch of set_influence_function, and it is a heavy addition for a machine that only builds reconstructors. Install it with the plot extra below; calling one of those methods without it raises an ImportError naming the extra.

Optional extras¶

Plotting

pip install "pyTomoAO[plot]"

Adds matplotlib, which the visualisation helpers import on demand.

GPU acceleration

pip install "pyTomoAO[gpu]"

which pulls in CuPy for CUDA 12. On CUDA 11, install the matching wheel yourself instead:

pip install cupy-cuda11x

pyTomoAO detects CuPy at import time and switches to the GPU covariance kernels automatically. If CuPy is installed but fails to load — a driver or toolkit mismatch, or no visible device — pyTomoAO logs a warning with the underlying error and falls back to the CPU backend, rather than silently reporting that CUDA is unavailable. See GPU acceleration.

Documentation toolchain

pip install ".[docs]"

Installs Sphinx, the Furo theme and the MyST/design extensions needed to build this site locally. See Documentation.

Verifying the installation¶

python -c "import pyTomoAO; print(pyTomoAO.__version__)"

Logging¶

pyTomoAO logs through the standard logging module and does not configure logging for you — importing it is silent. Turn its messages on from your own code:

import logging

logging.basicConfig(level=logging.INFO)

from pyTomoAO.reconstructor import tomographicReconstructor

You will then see progress messages, including which backend was selected:

INFO:pyTomoAO.reconstructor:
CUDA is not available. Using CPU for computations.

Both backends are fully supported — that message is informational, not an error.

To keep pyTomoAO quiet while your own application logs at INFO:

logging.getLogger("pyTomoAO").setLevel(logging.WARNING)

Next: Quickstart.