Add better diagnostics & plotting (#67)

* Add caching functions

* Add limts

* Add new mass runs

* Update .gitignore

* Edit which CDFs are loaded

* Stop saving cross hindxs

* Change dist to half precision

* New nearest path

* Add neighbour counting

* Add neighbour kwargs

* Update work in progress

* Add new counting

* Add how dist is built

* Collect dist to 1 file

* Update reading routine

* Delete Quijote files

* Remove file

* Back to float32

* Fix bugs

* Rename utils

* Remove neighbuor kwargs

* Rename file

* Fix bug

* Rename plt utils

* Change where nghb kwargs from

* line length

* Remove old notebooks

* Move survey

* Add white space

* Update TODO

* Update CDF calculation

* Update temporarily plotting

* Merge branch 'add_diagnostics' of github.com:Richard-Sti/csiborgtools into add_diagnostics

* Start adding documentation to plotting

* Remove comments

* Better code documentation

* Some work on tidal tensor

* Better plotting

* Add comment

* Remove nb

* Remove comment

* Add documentation

* Update plotting

* Update submission

* Update KL vs KS plots

* Update the plotting routine

* Update plotting

* Update plotting routines
This commit is contained in:
Richard Stiskalek 2023-06-16 14:33:27 +01:00 committed by GitHub
parent 004d9629a2
commit ccbbbd24b4
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
20 changed files with 1075 additions and 32121 deletions

View file

@ -19,12 +19,14 @@ MPI parallelized over the reference simulations.
from argparse import ArgumentParser
from datetime import datetime
from distutils.util import strtobool
from os import remove
import numpy
import yaml
from mpi4py import MPI
from taskmaster import work_delegation
from tqdm import trange
from utils import open_catalogues
try:
@ -36,7 +38,7 @@ except ModuleNotFoundError:
import csiborgtools
def find_neighbour(args, nsim, cats, paths, comm):
def find_neighbour(args, nsim, cats, paths, comm, save_kind):
"""
Find the nearest neighbour of each halo in the given catalogue.
@ -53,23 +55,78 @@ def find_neighbour(args, nsim, cats, paths, comm):
Paths object.
comm : mpi4py.MPI.Comm
MPI communicator.
save_kind : str
Kind of data to save. Must be either `dist` or `bin_dist`.
Returns
-------
None
"""
assert save_kind in ["dist", "bin_dist"]
ndist, cross_hindxs = csiborgtools.match.find_neighbour(nsim, cats)
mass_key = "totpartmass" if args.simname == "csiborg" else "group_mass"
cat0 = cats[nsim]
mass = cat0[mass_key]
rdist = cat0.radial_distance(in_initial=False)
fout = paths.cross_nearest(args.simname, args.run, nsim)
# Distance is saved optionally, whereas binned distance is always saved.
if save_kind == "dist":
out = {"ndist": ndist,
"cross_hindxs": cross_hindxs,
"mass": cat0[mass_key],
"ref_hindxs": cat0["index"],
"rdist": rdist}
fout = paths.cross_nearest(args.simname, args.run, "dist", nsim)
if args.verbose:
print(f"Rank {comm.Get_rank()} writing to `{fout}`.", flush=True)
numpy.savez(fout, **out)
paths = csiborgtools.read.Paths(**csiborgtools.paths_glamdring)
reader = csiborgtools.read.NearestNeighbourReader(
paths=paths, **csiborgtools.neighbour_kwargs)
counts = numpy.zeros((reader.nbins_radial, reader.nbins_neighbour),
dtype=numpy.float32)
counts = reader.count_neighbour(counts, ndist, rdist)
out = {"counts": counts}
fout = paths.cross_nearest(args.simname, args.run, "bin_dist", nsim)
if args.verbose:
print(f"Rank {comm.Get_rank()} writing to `{fout}`.", flush=True)
numpy.savez(fout, ndist=ndist, cross_hindxs=cross_hindxs, mass=mass,
ref_hindxs=cat0["index"], rdist=rdist)
numpy.savez(fout, **out)
def collect_dist(args, paths):
"""
Collect the binned nearest neighbour distances into a single file.
Parameters
----------
args : argparse.Namespace
Command line arguments.
paths : csiborgtools.paths.Paths
Paths object.
Returns
-------
"""
fnames = paths.cross_nearest(args.simname, args.run, "bin_dist")
if args.verbose:
print("Collecting counts into a single file.", flush=True)
for i in trange(len(fnames)) if args.verbose else range(len(fnames)):
fname = fnames[i]
data = numpy.load(fname)
if i == 0:
out = data["counts"]
else:
out += data["counts"]
remove(fname)
fout = paths.cross_nearest(args.simname, args.run, "tot_counts",
nsim=0, nobs=0)
if args.verbose:
print(f"Writing the summed counts to `{fout}`.", flush=True)
numpy.savez(fout, tot_counts=out)
if __name__ == "__main__":
@ -87,16 +144,23 @@ if __name__ == "__main__":
with open("./match_finsnap.yml", "r") as file:
config = yaml.safe_load(file)
if args.simname == "csiborg":
save_kind = "dist"
else:
save_kind = "bin_dist"
comm = MPI.COMM_WORLD
rank = comm.Get_rank()
paths = csiborgtools.read.Paths(**csiborgtools.paths_glamdring)
cats = open_catalogues(args, config, paths, comm)
def do_work(nsim):
return find_neighbour(args, nsim, cats, paths, comm)
return find_neighbour(args, nsim, cats, paths, comm, save_kind)
work_delegation(do_work, list(cats.keys()), comm,
master_verbose=args.verbose)
comm.Barrier()
if comm.Get_rank() == 0:
if rank == 0:
collect_dist(args, paths)
print(f"{datetime.now()}: all finished. Quitting.")

View file

@ -18,20 +18,77 @@ nbins_marks: 10
name:
- totpartmass
- group_mass
min: 1.e+12
max: 1.e+13
min: 12.4
max: 12.8
islog: true
"mass002":
primary:
name:
- totpartmass
- group_mass
min: 1.e+13
max: 1.e+14
min: 12.6
max: 13.0
islog: true
"mass003":
primary:
name:
- totpartmass
- group_mass
min: 1.e+14
min: 12.8
max: 13.2
islog: true
"mass004":
primary:
name:
- totpartmass
- group_mass
min: 13.0
max: 13.4
islog: true
"mass005":
primary:
name:
- totpartmass
- group_mass
min: 13.2
max: 13.6
islog: true
"mass006":
primary:
name:
- totpartmass
- group_mass
min: 13.4
max: 13.8
islog: true
"mass007":
primary:
name:
- totpartmass
- group_mass
min: 13.6
max: 14.0
islog: true
"mass008":
primary:
name:
- totpartmass
- group_mass
min: 13.8
max: 14.2
islog: true
"mass009":
primary:
name:
- totpartmass
- group_mass
min: 14.0
islog: true

View file

@ -106,8 +106,12 @@ def read_single_catalogue(args, config, nsim, run, rmax, paths, nobs=None):
pname = _name
if pname is None:
raise KeyError(f"Invalid names `{sel['name']}`.")
cat.apply_bounds({pname: (sel.get("min", None), sel.get("max", None))})
xmin = sel.get("min", None)
xmax = sel.get("max", None)
if sel.get("islog", False):
xmin = 10**xmin if xmin is not None else None
xmax = 10**xmax if xmax is not None else None
cat.apply_bounds({pname: (xmin, xmax)})
# Now the secondary selection bounds. If needed transfrom the secondary
# property before applying the bounds.