Skip to content

conda and mamba

This page changed after phonometry 4.0.0, the latest release on PyPI, so it may describe code that release does not have yet. What is new since then is listed under Unreleased in the changelog.

phonometry installs from PyPI. In a conda environment, take its dependencies from conda-forge first, then install phonometry with the environment’s own pip: pip finds every requirement already there and adds phonometry alone, so NumPy, SciPy and the rest stay conda-forge builds.

Checked on 17 November 2026 with phonometry 4.0.0 from PyPI, in the condaforge/miniforge3 image: conda 26.9.4 and mamba 2.9.0, Python 3.13.18 with NumPy 2.5.5 from conda-forge, and Python 3.14.10 with pytorch 2.14.1.

Terminal window
conda create -n acoustics -c conda-forge python=3.13 numpy scipy matplotlib-base numba reportlab svglib pysoundfile
conda activate acoustics
python -m pip install "phonometry[full]"

pip reports each requirement as already satisfied and installs phonometry only. conda list then shows phonometry with the channel pypi beside the conda-forge packages, and python -m pip check reports nothing broken. mamba takes the same arguments: mamba create -n acoustics -c conda-forge ... builds the same environment. Python 3.12, 3.13 and 3.14 all work. conda-forge publishes no Python 3.12 or newer for 32-bit Windows or 32-bit Linux, so phonometry installs there with pip alone, [plot,report,audio] in place of [full], as pip and uv says.

Extra conda-forge packages
the base install numpy, scipy
[plot] matplotlib-base
[report] reportlab, svglib, and matplotlib-base for the fiche’s figure
[audio] pysoundfile, which brings libsndfile
[perf] numba
[torch] pytorch, array-api-compat

matplotlib-base is matplotlib without a GUI backend, which is what the library draws with; pysoundfile is the conda-forge name of the package PyPI calls soundfile. Install the conda-forge packages of the extras you want before pip runs. A dependency still missing when pip runs comes from PyPI instead, and conda list shows it with the channel pypi beside the conda-forge NumPy.

A feature whose dependency is missing names both spellings:

ImportError: Plotting requires matplotlib. Install it with: pip install "phonometry[plot]" (conda: conda install -c conda-forge matplotlib-base)

In a conda environment, the second one is the command to run.

On a machine without a GPU, conda-forge installs its CPU build of pytorch, so no index has to be named:

Terminal window
conda create -n fdtd -c conda-forge python=3.14 numpy scipy pytorch array-api-compat
conda activate fdtd
python -m pip install "phonometry[torch]"

pip again installs phonometry alone, and an FDTD run on Compute(device="torch:cpu") gives the same pressures as the NumPy run.