Payu is currently only supported for users on the NCI computing systems. If you wish to use payu on other systems, see the notes at the end of this document.
Payu is made available for users of NCI HPC systems in conda environments.
The ACCESS-Hive ACCESS-OM models documentation contains instructions for using ACCESS-NRI supported conda environments.
Using pip it is possible to install payu from PyPI:
pip install payu --user
If you want to use the latest version of payu, then you can install directly from the repository:
pip install payu@git+https://github.com/payu-org/payu --user
or clone the codebase and install from there:
git clone https://github.com/payu-org/payu
pip install . --user
Payu is not supported for general use, and it would be a tremendous surprise if it even worked on other machines. In particular, the following services are presumed to be available:
Environment Modules: Not only do we assume support for environment modules, but we also assume the existence of certain modules, such as an OpenMPI module and particular versions of Python.
PBS Scheduler: Payu relies on executables that are provided with most PBS implementations, such as Torque or PBSPro. Most of the argument flags are currently based around PBSPro conventions.
MPI: Jobs are submitted via
mpirunand most of the argument flags are based on the OpenMPI implementation.
There are also some additional assumptions based on the architecture of the NCI HPC facilities.
Despite these rather strict requirements, there is opportunity for generalising payu for other platforms, such as through new drivers for alternative schedulers and parallelisation platforms. Please create a GitHub Issue if you are interested in porting payu to your machine.