Closed yymao closed 3 years ago
On cori login node with SCRATCH file system, converting on one healpix of cosmoDC2 takes about 11 minutes, resulting in a 25 GB parquet file (per healpix, not including native quantities).
Note of self: need to prevent duplicated columns when all column names are in lower case.
Thanks @JoanneBogart (and also thanks to @plaszczy and @cwwalter for checking the output files off github).
This PR adds an
--healpix
option inscripts/write_gcr_to_parquet.py
so that it is more convenient to use this script to convert cosmoDC2 to parquet. The option can be used as:As a test, I am generating a few healpix pixels in
/global/cscratch1/sd/yymao/desc/cosmodc2-parquet
on NERSC if anyone wants to take a look. (cc @JoanneBogart @plaszczy @cwwalter)