def run_mcmc(igal):
fmcmc = os.path.join('/global/cscratch1/sd/chahah/provabgs/svda/healpix/',
str(hpix),
'provabgs.%i.hdf5' % meta['TARGETID'][igal])
if os.path.isfile(fmcmc):
# don't overwrite
return None
# get observations
# set prior
prior = Infer.load_priors([
Infer.UniformPrior(7., 12.5, label='sed'),
Infer.FlatDirichletPrior(4, label='sed'), # flat dirichilet priors
Infer.UniformPrior(0., 1., label='sed'), # burst fraction
Infer.UniformPrior(1e-2, 13.27, label='sed'), # tburst
Infer.LogUniformPrior(4.5e-5, 1.5e-2,
label='sed'), # log uniform priors on ZH coeff
Infer.LogUniformPrior(4.5e-5, 1.5e-2,
label='sed'), # log uniform priors on ZH coeff
Infer.UniformPrior(0., 3., label='sed'), # uniform priors on dust1
Infer.UniformPrior(0., 3., label='sed'), # uniform priors on dust2
Infer.UniformPrior(-2., 1.,
label='sed'), # uniform priors on dust_index
Infer.GaussianPrior(f_fiber[igal],
sigma_f_fiber[igal]**2,
label='flux_calib') # flux calibration
])
desi_mcmc = Infer.desiMCMC(model=m_nmf,
prior=prior,
flux_calib=m_fluxcalib)
photo_flux_i = np.array(list(photo_flux[igal]))
photo_ivar_i = np.array(list(photo_ivar[igal]))
# run MCMC
zeus_chain = desi_mcmc.run(
wave_obs=w_obs,
flux_obs=f_obs[igal, :],
flux_ivar_obs=i_obs[igal, :],
bands='desi', # g, r, z
photo_obs=photo_flux_i,
photo_ivar_obs=photo_ivar_i,
zred=zred[igal],
vdisp=0.,
sampler='zeus',
nwalkers=30,
burnin=0,
opt_maxiter=2000,
niter=niter,
progress=False,
debug=True,
writeout=fmcmc,
overwrite=True)
return None
def prior_burst():
''' prior on burst contribution
'''
return Infer.load_priors([
Infer.LogUniformPrior(
1e-2,
13.27), # log uniform priors on tburst from 10Myr to 13.27 Gyr
Infer.LogUniformPrior(4.5e-5, 4.5e-2,
label='sed'), # log uniform priors on Z burst
Infer.UniformPrior(0., 3., label='sed'), # uniform priors on dust1
Infer.UniformPrior(0., 3., label='sed'), # uniform priors on dust2
Infer.UniformPrior(-3., 1.,
label='sed') # uniform priors on dust_index
])
def prior_nmf(ncomp):
''' prior on 4 component NMF by Rita
'''
return Infer.load_priors([
Infer.FlatDirichletPrior(ncomp, label='sed'), # flat dirichilet priors
Infer.LogUniformPrior(4.5e-5, 4.5e-2,
label='sed'), # log uniform priors on ZH coeff
Infer.LogUniformPrior(4.5e-5, 4.5e-2,
label='sed'), # log uniform priors on ZH coeff
Infer.UniformPrior(0., 3., label='sed'), # uniform priors on dust1
Infer.UniformPrior(0., 3., label='sed'), # uniform priors on dust2
Infer.UniformPrior(-3., 1.,
label='sed'), # uniform priors on dust_index
Infer.UniformPrior(0., 0.6, label='sed') # uniformly sample redshift
])
def run_mcmc(igal):
# get observations
zred_i, photo_flux_i, photo_ivar_i, w_obs, f_obs, i_obs, f_fiber, sigma_f_fiber\
= sv.get_spectrophotometry(igal, sample=sample)
# set prior
prior = Infer.load_priors([
Infer.UniformPrior(7., 12.5, label='sed'),
Infer.FlatDirichletPrior(4, label='sed'), # flat dirichilet priors
Infer.UniformPrior(0., 1., label='sed'), # burst fraction
Infer.UniformPrior(1e-2, 13.27, label='sed'), # tburst
Infer.LogUniformPrior(4.5e-5, 1.5e-2,
label='sed'), # log uniform priors on ZH coeff
Infer.LogUniformPrior(4.5e-5, 1.5e-2,
label='sed'), # log uniform priors on ZH coeff
Infer.UniformPrior(0., 3., label='sed'), # uniform priors on dust1
Infer.UniformPrior(0., 3., label='sed'), # uniform priors on dust2
Infer.UniformPrior(-2., 1.,
label='sed'), # uniform priors on dust_index
Infer.GaussianPrior(f_fiber, sigma_f_fiber**2,
label='flux_calib') # flux calibration
])
desi_mcmc = Infer.desiMCMC(model=m_nmf,
prior=prior,
flux_calib=m_fluxcalib)
fmcmc = os.path.join('/global/cscratch1/sd/chahah/provabgs/raga/',
sample.replace('.fits', '.%i.hdf5' % igal))
# run MCMC
zeus_chain = desi_mcmc.run(
wave_obs=w_obs,
flux_obs=f_obs,
flux_ivar_obs=i_obs,
bands='desi', # g, r, z
photo_obs=photo_flux_i,
photo_ivar_obs=photo_ivar_i,
zred=zred_i,
vdisp=0.,
sampler='zeus',
nwalkers=30,
burnin=0,
opt_maxiter=2000,
niter=niter,
progress=True,
debug=True,
writeout=fmcmc,
overwrite=True)
return None
ibatch = sys.argv[1]
assert ibatch == 'test'
ncpu = int(sys.argv[2])
# hardcoded to NERSC directory for LRG
#dat_dir='/global/cscratch1/sd/chahah/provabgs/emulator' # hardcoded to NERSC directory
# for LRG
dat_dir='/global/cscratch1/sd/chahah/provabgs/emulator/lrg/'
###########################################################################################
# priors of burst component
priors = Infer.load_priors([
Infer.FlatDirichletPrior(4, label='sed'), # flat dirichilet priors
Infer.LogUniformPrior(4.5e-5, 2.0e-2, label='sed'), # log uniform priors on ZH coeff
Infer.LogUniformPrior(4.5e-5, 2.0e-2, label='sed'), # log uniform priors on ZH coeff
Infer.UniformPrior(0., 3., label='sed'), # uniform priors on dust1
Infer.UniformPrior(0., 3., label='sed'), # uniform priors on dust2
Infer.UniformPrior(-3., 1., label='sed'), # uniform priors on dust_index
Infer.UniformPrior(0.3, 1.5, label='sed') # uniformly sample redshift range of LRG
])
# redshift range for BGS
# Infer.UniformPrior(0., 0.6, label='sed') # uniformly sample redshift
if ibatch == 'test':
np.random.seed(123456)
nspec = 100000 # batch size
ftheta = os.path.join(dat_dir, 'fsps.%s.v%s.theta.test.npy' % (name, version))
ftheta_unt = os.path.join(dat_dir, 'fsps.%s.v%s.theta_unt.test.npy' % (name, version))
fspectrum = os.path.join(dat_dir, 'fsps.%s.v%s.lnspectrum.test.npy' % (name, version))
else:
np.random.seed(ibatch)
def multiprocessing_zeus():
'''
'''
#
fsps_emulator = Models.DESIspeculator()
# set prior
priors = Infer.load_priors([
Infer.UniformPrior(10., 10.5, label='sed'),
Infer.FlatDirichletPrior(4, label='sed'),
Infer.UniformPrior(np.array([6.9e-5, 6.9e-5, 0., 0., -2.2]), np.array([7.3e-3, 7.3e-3, 3., 4., 0.4]), label='sed')
])
random_theta = priors.sample()
wave, flux = fsps_emulator.sed(priors.transform(random_theta), 0.1)
desi_mcmc = Infer.desiMCMC(prior=priors)
t0 = time.time()
mcmc = desi_mcmc.run(
wave_obs=wave[0],
flux_obs=flux[0],
flux_ivar_obs=np.ones(flux.shape[1]),
zred=0.1,
sampler='zeus',
nwalkers=20,
burnin=10,
opt_maxiter=1000,
niter=100,
pool=None,
debug=True)
print()
print('running on series takes %.f' % (time.time() - t0))
print()
import zeus
import multiprocessing
ncpu = multiprocessing.cpu_count()
print('%i cpus' % ncpu)
t0 = time.time()
lnpost_args, lnpost_kwargs = desi_mcmc._lnPost_args_kwargs(
wave_obs=wave[0],
flux_obs=flux[0],
flux_ivar_obs=np.ones(flux.shape[1]),
zred=0.1)
start = desi_mcmc._initialize_walkers(lnpost_args, lnpost_kwargs, priors,
nwalkers=20, opt_maxiter=1000, debug=True)
print('--- burn-in ---')
pewl = Pool(processes=ncpu)
with pewl as pool:
zeus_sampler = zeus.EnsembleSampler(
desi_mcmc.nwalkers,
desi_mcmc.prior.ndim,
desi_mcmc.lnPost,
pool=pool,
args=lnpost_args,
kwargs=lnpost_kwargs)
zeus_sampler.run_mcmc(start, 10)
burnin = zeus_sampler.get_chain()
print('--- running main MCMC ---')
pewl = Pool(processes=ncpu)
with pewl as pool:
zeus_sampler = zeus.EnsembleSampler(
desi_mcmc.nwalkers,
desi_mcmc.prior.ndim,
desi_mcmc.lnPost,
pool=pool,
args=lnpost_args,
kwargs=lnpost_kwargs)
zeus_sampler.run_mcmc(burnin[-1], 100)
_chain = zeus_sampler.get_chain()
print()
print('running on parallel takes %.f' % (time.time() - t0))
print()
return None
###########################################################################################
name = 'burst'
version = '0.1'
try:
ibatch = int(sys.argv[1])
except ValueError:
ibatch = sys.argv[1]
assert ibatch == 'test'
ncpu = int(sys.argv[2])
###########################################################################################
# priors of burst component
priors = Infer.load_priors([
Infer.UniformPrior(
1e-2, 13.27), # uniform priors on tburst from 10Myr to 13.27 Gyr
Infer.LogUniformPrior(4.5e-5, 4.5e-2,
label='sed'), # log uniform priors on Z burst
Infer.UniformPrior(0., 3., label='sed'), # uniform priors on tau_ISM
Infer.UniformPrior(-3., 1., label='sed') # uniform priors on dust_index
])
dat_dir = '/global/cscratch1/sd/chahah/provabgs/emulator' # hardcoded to NERSC directory
if ibatch == 'test':
np.random.seed(123456)
nspec = 100000 # batch size
ftheta = os.path.join(dat_dir,
'fsps.%s.v%s.theta.test.npy' % (name, version))
fspectrum = os.path.join(
dat_dir, 'fsps.%s.v%s.lnspectrum.test.npy' % (name, version))
else:
np.random.seed(ibatch)
nspec = 10000 # batch size