Runtime options - GabrielaCR/AGNfitter GitHub Wiki
Runtime options
You can see them with python RUN_AGNfitter_multi.py -h
:
XXXX
___________ XXX _________________________________________________
XX
X
X AGNfitter
__ X __ ---------
/**\ | /**\
... (*** = o = ***) ...........................................
\**/__ | __\**/
X Fitting SEDs of AGN and Galaxies
X in a MCMC Approach
xx (Calistro Rivera et al. 2016)
xx
_______ xxx______________________________________________________
xxxx
usage: RUN_AGNfitter_multi.py [-h] [-c NCPU] [-n SOURCENUMBER] [-i] [-o]
AGNfitterSettings
positional arguments:
AGNfitterSettings AGNfitter settings file
optional arguments:
-h, --help show this help message and exit
-c NCPU, --ncpu NCPU number of cpus to use for multiprocessing
-n SOURCENUMBER, --sourcenumber SOURCENUMBER
specify a single source number to run (this is the
line number in hte catalogue not the source id/name)
-i, --independent run independently per source, i.e. do not create a
global model dictionary
-o, --overwrite overwrite model files
Single source
To run AGNfitter for a single source in the catalog, specify the line number as the sourcenumber argument: e.g.
./RUN_AGNfitter_multi.py --sourcenumber 0 my_SETTINGS_AGNfitter.py
Working on a computer with multiple cores
To run AGNfitter in batch mode using python's multiprocessing capability and improve the efficiency, run e.g on a machine with 10 cpu cores
./RUN_AGNfitter_multi.py --ncpu 10 my_SETTINGS_AGNfitter.py
Working on a computer cluster with multiple different machines
To run AGNfitter in a distributed mode on a compute cluster with multiple machines, a shared filesystem and a queue system, e.g using a PBS array job to specify the calalog line numbers
./RUN_AGNfitter_multi.py --independent --sourcenumber $PBS_ARRAY_ID my_SETTINGS_AGNfitter.py
The --independent
flag is required so that each job produces its own model dictionary at its own redshift (i.e. each machine does not recreate the model dictionaries for the entire catalog). This can be more efficient for large catalogs where the model dictionary creation (which is not paralellized) can take a long time.
Additionally, you can specify --overwrite
if you wish to recreate any existing models dictionaries (in case you change the z arrays).