A Python interface to the EZmock code. This package does two things:
- Automate the process of creating EZmocks on an HPC cluster managed with Slurm.
- Easily load created EZmocks as Python objects for convenient manipulation in (say) Jupyter notebooks.
You should already have a compiled EZmock binary somewhere. Make sure you have execute permissions for the binary.
You should also have a environment variable $SCRATCH defined on your system,
giving the path to the your scratch directory.
Then clone this repo,
enter the conda environment you want to install inside,
and install via pip:
pip install -e .This step should install all necessary dependencies.
Next, edit the hardcoded path EZMOCK_BINARY_PATH in ezmock/ezmock_job.py
to point to your EZmock binary.
If you want to compute bispectra with this code,
build Cheng Zhao's bispec script and edit BISPEC_BINARY in ezmock/ezmock_job.py.
Then, create a Slurm batch script template ezmock-multisubmit-template.sbatch
in the templates directory to fit your Slurm needs.
There are some examples in the directory; you should basically only need to modify the slurm options.
Make sure to edit the following options:
--outputand--errorshould point to your desired Slurm output paths--cpus-per-task(on NERSC, for example, it's often better to have this be twice the number of OpenMP threads)--partition--accountOMP_NUM_THREADS
ezmock will require a $SCRATCH/ezmock directory,
where it stores all relevant files,
including EZmock catalogs.
Please refer to the example notebook (docs/tutorial.ipynb).
Pull requests are most welcome, as this code grew organically from some personal scripts and is by no means perfect.