Source code for the BNL GPU Hackathon
- https://portal.nersc.gov/project/hacc/aphearin/BNLHack19/
- https://drive.google.com/file/d/1JZn3fbBELazpnrlw-0WiGzBkNZaFlT_7/view?usp=sharing
- The code in the
sf_history
subpackage maps galaxy properties onto halos. For example, themean_sfr
function maps SFR onto halos across time; this function needs to be integrated to map stellar mass onto halos. - The code in the
mock_obs
subpackage computes summary statistics of the galaxy population. Thenumba_gw_hist
function calculates a Gaussian-weighted histogram, the kernel underlying a stellar mass function.
-
Install miniconda.
I've been told it is better to keep it out of your home area, but that might just be my account. I made a directory on the global scratch area, symlinked it to my home area, and then installed there.
The global scratch area is
/hpcgpfs01/scratch
. -
Use
conda init
to setup your shell if you did not install it while doing miniconda. -
Add
module load "cuda/9.0"
to your~/.bashrc
. -
Install all of the software you need in your base conda env (e.g.,
numba
,pytest
,numpy
,scipy
,jax
, etc.). -
Get a node
srun --pty -A hackathon -N 1 --exclusive -p long -t 3:00:00 /bin/bash