Installation#
luas
requires jax
to run so see the jax installation page
for instructions on how to install it for your specific choice of hardware. A GPU is not required to run any of the code but it can significantly speed up many of the calculations so be sure to follow the instructions for installing with GPU if you have one available.
Note
Due to the required numerical precision for matrix decomposition, double-precision floating point arithmetic is required which many GPUs either do not support or will perform poorly at. Make sure to check the specific hardware capabilities of the GPU being used e.g. number of double-precision FLOPS. GPUs which will perform exceptionally well include common HPC hardware such as NVIDIA TESLA V100/A100s.
luas
also does not have its own inference library but has custom distributions for NumPyro
and PyMC
so for most purposes if you want to perform inference it is likely you will need to install one of these. See the installation guides for NumPyro
and PyMC
. The latest versions are ideal but luas
is compatible with PyMC
v4 as well as the current PyMC
v5.
If performing exoplanet transit modelling with the luas.exoplanet
submodule then you will need the additional dependency jaxoplanet
which should be installable using pip (see jaxoplanet
for the latest information). Currently luas
only supports jaxoplanet
v0.0.1 although we hope to update to the more recent version soon.
Installation of luas
can be performed by cloning the GitHub repository and using pip.
git clone https://github.com/markfortune/luas.git
cd luas
python -m pip install -e .