def test_neff():
# test of neff functionality
ndim = 2
rstate = get_rstate()
sampler = dynesty.NestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
rstate=rstate)
assert sampler.n_effective == 0
sampler.run_nested(print_progress=printing)
assert sampler.n_effective > 10
sampler = dynesty.DynamicNestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
rstate=rstate)
assert sampler.n_effective == 0
sampler.run_nested(dlogz_init=1, n_effective=1000, print_progress=printing)
assert sampler.n_effective > 1000
sampler = dynesty.DynamicNestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
rstate=rstate)
sampler.run_nested(dlogz_init=1,
n_effective=10000,
print_progress=printing)
assert sampler.n_effective > 10000
def __init__(self,
likelihood_module,
prior_type='uniform',
prior_means=None,
prior_sigmas=None,
width_scale=1,
sigma_scale=1,
bound='multi',
sample='auto',
use_mpi=False,
use_pool={}):
"""
:param likelihood_module: likelihood_module like in likelihood.py (should be callable)
:param prior_type: 'uniform' of 'gaussian', for converting the unit hypercube to param cube
:param prior_means: if prior_type is 'gaussian', mean for each param
:param prior_sigmas: if prior_type is 'gaussian', std dev for each param
:param width_scale: scale the widths of the parameters space by this factor
:param sigma_scale: if prior_type is 'gaussian', scale the gaussian sigma by this factor
:param bound: specific to Dynesty, see https://dynesty.readthedocs.io
:param sample: specific to Dynesty, see https://dynesty.readthedocs.io
:param use_mpi: Use MPI computing if `True`
:param use_pool: specific to Dynesty, see https://dynesty.readthedocs.io
"""
super(DynestySampler,
self).__init__(likelihood_module, prior_type, prior_means,
prior_sigmas, width_scale, sigma_scale)
# create the Dynesty sampler
if use_mpi:
from schwimmbad import MPIPool
import sys
pool = MPIPool(use_dill=True)
if not pool.is_master():
pool.wait()
sys.exit(0)
self._sampler = dynesty.DynamicNestedSampler(self.log_likelihood,
self.prior,
self.n_dims,
bound=bound,
sample=sample,
pool=pool,
use_pool=use_pool)
else:
self._sampler = dynesty.DynamicNestedSampler(self.log_likelihood,
self.prior,
self.n_dims,
bound=bound,
sample=sample)
self._has_warned = False
def __init__(self, model, nlive, nprocesses=1,
loglikelihood_function=None, use_mpi=False, run_kwds=None,
**kwargs):
self.model = model
log_likelihood_call, prior_call = setup_calls(
model,
nprocesses=nprocesses,
loglikelihood_function=loglikelihood_function)
# Set up the pool
pool = choose_pool(mpi=use_mpi, processes=nprocesses)
if pool is not None:
pool.size = nprocesses
self.run_kwds = {} if run_kwds is None else run_kwds
self.nlive = nlive
self.names = model.sampling_params
self.ndim = len(model.sampling_params)
self.checkpoint_file = None
if self.nlive < 0:
# Interpret a negative input value for the number of live points
# (which is clearly an invalid input in all senses)
# as the desire to dynamically determine that number
self._sampler = dynesty.DynamicNestedSampler(log_likelihood_call,
prior_call, self.ndim,
pool=pool, **kwargs)
else:
self._sampler = dynesty.NestedSampler(log_likelihood_call,
prior_call, self.ndim,
nlive=self.nlive,
pool=pool, **kwargs)
def test_results(dyn):
# test of various results interfaces functionality
ndim = 2
rstate = get_rstate()
if dyn:
sampler = dynesty.DynamicNestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
rstate=rstate)
else:
sampler = dynesty.NestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
rstate=rstate)
sampler.run_nested(print_progress=printing)
res = sampler.results
for k in res.keys():
pass
for k, v in res.items():
pass
for k, v in res.asdict().items():
pass
print(res)
print(str(res))
print('logl' in res)
res1 = res.copy()
# check it's pickleable
S = pickle.dumps(res)
res = pickle.loads(S)
def test_periodic():
# hard test of dynamic sampler with high dlogz_init and small number
# of live points
logz_true = np.log(np.sqrt(2 * np.pi) * erf(win / np.sqrt(2)) / (2 * win))
thresh = 5
ndim = 2
rstate = get_rstate()
sampler = dynesty.DynamicNestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
periodic=[0],
rstate=rstate)
sampler.run_nested(dlogz_init=1, print_progress=printing)
assert (np.abs(sampler.results.logz[-1] - logz_true) <
thresh * sampler.results.logzerr[-1])
sampler = dynesty.NestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
periodic=[0],
rstate=rstate)
sampler.run_nested(dlogz=1, print_progress=printing)
assert (np.abs(sampler.results.logz[-1] - logz_true) <
thresh * sampler.results.logzerr[-1])
def __init__(self, loglikelihood,
prior_transform, ndim,
sample='auto', bound='multi',
n_cpu=None, n_thread=None):
if n_cpu is None:
n_cpu = mp.cpu_count()
if n_thread is not None:
n_thread = max(n_thread, n_cpu-1)
if n_cpu > 1:
self.open_pool(n_cpu)
self.use_pool = {'update_bound': False}
else:
self.pool = None
self.use_pool = None
self.prior_tf = prior_transform
self.loglike = loglikelihood
self.ndim = ndim
dsampler = dynesty.DynamicNestedSampler(self.loglike,
self.prior_tf,
self.ndim,
sample=sample,
bound=bound,
pool=self.pool,
queue_size=n_thread,
use_pool=self.use_pool)
self.dsampler = dsampler
def test_pickle(dynamic, with_pool):
# test of pickling functionality
ndim = 2
rstate = get_rstate()
if with_pool:
kw = dict(pool=Pool(2), queue_size=100)
else:
kw = {}
if dynamic:
sampler = dynesty.DynamicNestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
rstate=rstate,
**kw)
else:
sampler = dynesty.NestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
rstate=rstate,
**kw)
sampler.run_nested(print_progress=printing, maxiter=100)
# i do it twice as there were issues previously
# with incorrect pool restoring
S = pickle.dumps(sampler)
sampler = pickle.loads(S)
S = pickle.dumps(sampler)
sampler = pickle.loads(S)
sampler.run_nested(print_progress=printing, maxiter=100)
if with_pool:
kw['pool'].close()
kw['pool'].join()
def test_periodic(sampler, dynamic):
# hard test of dynamic sampler with high dlogz_init and small number
# of live points
logz_true = np.log(np.sqrt(2 * np.pi) * erf(win / np.sqrt(2)) / (2 * win))
thresh = 8
# This is set up to higher level
# becasue of failures at ~5ssigma level
# this needs to be investigated
rstate = get_rstate()
if dynamic:
dns = dynesty.DynamicNestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
periodic=[0],
rstate=rstate,
sample=sampler)
dns.run_nested(dlogz_init=1, print_progress=printing)
else:
dns = dynesty.NestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
periodic=[0],
rstate=rstate,
sample=sampler)
dns.run_nested(dlogz=1, print_progress=printing)
assert (np.abs(dns.results.logz[-1] - logz_true) <
thresh * dns.results.logzerr[-1])
def test_periodic(sampler, dynamic):
# hard test of dynamic sampler with high dlogz_init and small number
# of live points
logz_true = np.log(np.sqrt(2 * np.pi) * erf(win / np.sqrt(2)) / (2 * win))
thresh = 5
rstate = get_rstate()
if dynamic:
dns = dynesty.DynamicNestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
sample=sampler,
reflective=[0],
rstate=rstate)
else:
dns = dynesty.NestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
sample=sampler,
reflective=[0],
rstate=rstate)
dns.run_nested(print_progress=printing)
assert (np.abs(dns.results.logz[-1] - logz_true) <
thresh * dns.results.logzerr[-1])
def run_dynamic_nested(self,
evidence=False,
posterior=False,
nlive=250,
sample='slice'):
self.sampler = dynesty.DynamicNestedSampler(
self.model.lnlikefunc,
self.model.priors.transform_prior,
ndim=self.ndim,
bound='multi',
sample=sample,
nlive=nlive)
if evidence and posterior:
print('Both, really? Going to default')
self.sampler.run_nested()
elif evidence:
# evidence focused dynamic run
self.sampler.run_nested(wt_kwargs={'pfrac': 0.0},
stop_kwargs={'pfrac': 0.0})
elif posterior:
# evidence focused dynamic run
self.sampler.run_nested(wt_kwargs={'pfrac': 1.0})
else:
# Default behavior, 80/20 weight split and 100% posterior.
self.sampler.run_nested()
def test_error():
rstate = get_rstate()
with pytest.raises(ValueError):
dynesty.DynamicNestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
reflective=[22],
rstate=rstate)
def test_dynamic():
# check dynamic nested sampling behavior
logz_tol = 1
dsampler = dynesty.DynamicNestedSampler(loglikelihood_gau,
prior_transform_gau,
ntotdim,
ncdim=ndim_gau)
dsampler.run_nested(print_progress=printing)
check_results_gau(dsampler.results, logz_tol)
def fit(self,
log_likelihood,
start,
num_dim,
prior_transform,
save_dims=None,
uid=None):
import dynesty
filename = self.get_filename(uid)
if os.path.exists(filename):
self.logger.info("Not sampling, returning result from file.")
return self.load_file(filename)
self.logger.info("Sampling posterior now")
if save_dims is None:
save_dims = num_dim
self.logger.debug("Fitting framework with %d dimensions" % num_dim)
self.logger.info("Using dynesty Sampler")
if self.dynamic:
sampler = dynesty.DynamicNestedSampler(log_likelihood,
prior_transform, num_dim)
sampler.run_nested(maxiter=self.max_iter,
print_progress=False,
nlive_init=self.nlive,
nlive_batch=100,
maxbatch=10)
else:
sampler = dynesty.NestedSampler(log_likelihood,
prior_transform,
num_dim,
nlive=self.nlive)
sampler.run_nested(maxiter=self.max_iter, print_progress=False)
self.logger.debug("Fit finished")
dresults = sampler.results
logz = dresults["logz"]
chain = dresults["samples"]
weights = np.exp(dresults["logwt"] - dresults["logz"][-1])
max_weight = weights.max()
trim = max_weight / 1e5
mask = weights > trim
likelihood = dresults["logl"]
self._save(chain[mask, :], weights[mask], likelihood[mask], filename,
logz[mask], save_dims)
return {
"chain": chain[mask, :],
"weights": weights[mask],
"posterior": likelihood[mask],
"evidence": logz
}
def test_stop_nmc():
# test stopping relying in n_mc
ndim = 2
rstate = get_rstate()
sampler = dynesty.DynamicNestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
rstate=rstate)
sampler.run_nested(dlogz_init=1,
n_effective=None,
stop_kwargs=dict(n_mc=25),
print_progress=printing)
def Model_2_sampler(prior_xform2,data1,data2,bins_,label):
x,n1,n2,dn1,dn2 = bin_data(data1,data2,bins_,label)
print("running the nested sampler... this might take from minutes to hours...")
dsampler = dynesty.DynamicNestedSampler(logLjoint2_skew, prior_xform2, ndim=16,
logl_args=(n1, n2, x),
nlive=2000,
bound='multi',
sample='auto')
dsampler.run_nested()
dres2 = dsampler.results
with open('sampler_results_model2_'+label, 'wb') as dres2_file:
pickle.dump(dres2, dres2_file)
print("sampler output saved as pickle file 'sampler_results_model2_"+label+"'")
def _fit_impl_impl(self, objective, parameters):
ndim = 3
sampler = dynesty.DynamicNestedSampler(_log_likelihood_wrapper,
_prior_tansform_wrapper,
ndim,
logl_args=(objective,
parameters),
ptform_args=(),
**self._options_constructor)
print(self._options_run_nested)
result = sampler.run_nested(**self._options_run_nested)
return result
def test_oldstop():
# test of old stopping function functionality
ndim = 2
rstate = get_rstate()
import dynesty.utils as dyutil
stopfn = dyutil.old_stopping_function
sampler = dynesty.DynamicNestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
rstate=rstate)
sampler.run_nested(dlogz_init=1,
n_effective=None,
stop_function=stopfn,
print_progress=printing)
def test_pool_dynamic():
# test pool in dynamic mode
# here for speed I do a gaussian
rstate = get_rstate()
with mp.Pool(2) as pool:
sampler = dynesty.DynamicNestedSampler(loglike_gau,
prior_transform_gau,
ndim,
nlive=nlive,
pool=pool,
queue_size=100,
rstate=rstate)
sampler.run_nested(dlogz_init=1, print_progress=printing)
assert (abs(LOGZ_TRUTH_GAU - sampler.results.logz[-1]) <
5. * sampler.results.logzerr[-1])
def test_maxcall():
# hard test of dynamic sampler with high dlogz_init and small number
# of live points
ndim = 2
sampler = dynesty.NestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive)
sampler.run_nested(dlogz=1, maxcall=1000)
sampler = dynesty.DynamicNestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive)
sampler.run_nested(dlogz_init=1, maxcall=1000)
def test_dyn():
# hard test of dynamic sampler with high dlogz_init and small number
# of live points
ndim = 2
bound = 'multi'
sampler = dynesty.DynamicNestedSampler(loglike_egg,
prior_transform_egg,
ndim,
nlive=nlive,
bound=bound,
sample='unif')
sampler.run_nested(dlogz_init=1, print_progress=printing)
logz_truth = 235.856
assert (abs(logz_truth - sampler.results.logz[-1]) <
5. * sampler.results.logzerr[-1])
def fit_e_disk2():
pdfs_weights_file = 'all_pdfs_weights.pkl'
membership_file = 'membership_probs.fits'
tmp = load_pdfs_weights_pickle(pdfs_weights_file)
pdf_dict = tmp[0]
wgt_dict = tmp[1]
d1_dict = tmp[2]
d2_dict = tmp[3]
grp_dict = tmp[4]
prob_mem = Table.read(membership_file)
# Fit only stars with non-zero membership probability.
p_thresh = 0.1
s_d2 = np.where(prob_mem['p_d2'] > p_thresh)[0]
e_solver = Eccentricity_Solver(pdf_dict['e'][s_d2, :],
wgt_dict['d2'][s_d2, :],
prob_mem['name'][s_d2])
e_solver.priors['alpha'] = make_gen(0, 50)
e_solver.priors['beta'] = make_gen(0, 30)
t0 = time.time()
# n_cpu = 4
# pool = Pool(n_cpu)
sampler = dynesty.DynamicNestedSampler(e_solver.LogLikelihood,
e_solver.Prior,
ndim=e_solver.n_dims,
bound='multi',
sample='unif')
sampler.run_nested(print_progress=True,
dlogz_init=0.05,
nlive_init=1000,
nlive_batch=500,
maxiter_init=20000,
maxiter_batch=2000,
maxbatch=10)
e_solver.sampler = sampler
t1 = time.time()
print('Runtime: ', t1 - t0)
e_solver.save('dnest_ecc_d2.pkl')
return
def test_pool2():
# test pool
ndim = 2
pool = mp.Pool(2)
sampler = dynesty.DynamicNestedSampler(loglike_egg,
prior_transform_egg,
ndim,
nlive=nlive,
bound='multi',
sample='unif',
pool=pool,
queue_size=2)
sampler.run_nested(dlogz_init=0.1, print_progress=printing)
logz_truth = 235.856
assert (abs(logz_truth - sampler.results.logz[-1]) <
5. * sampler.results.logzerr[-1])
def test_printing():
# hard test of dynamic sampler with high dlogz_init and small number
# of live points
ndim = 2
sampler = dynesty.DynamicNestedSampler(
loglike,
prior_transform,
ndim,
nlive=nlive,
)
sampler.run_nested(dlogz_init=1, print_progress=printing)
sampler = dynesty.NestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive)
sampler.run_nested(dlogz=1, print_progress=printing)
def test_dyn():
# hard test of dynamic sampler with high dlogz_init and small number
# of live points
ndim = 2
THRESHOLD = 5 # in sigmas
rstate = get_rstate()
# this is expected to use unif sampler and multi bound
sampler = dynesty.DynamicNestedSampler(loglike_egg,
prior_transform_egg,
ndim,
nlive=nlive,
rstate=rstate)
sampler.run_nested(dlogz_init=1, print_progress=printing)
assert (abs(LOGZ_TRUTH - sampler.results.logz[-1]) <
THRESHOLD * sampler.results.logzerr[-1])
print(sampler.citations)
def Run_Dynamic_Nested_Fitting(loglikelihood,
prior_transform, ndim,
nlive_init=100, sample='auto',
nlive_batch=50, maxbatch=2,
pfrac=0.8, n_cpu=None,
print_progress=True):
""" Run Fitting as a Function.
Parameters
----------
loglikelihood: function
log likelihood function
prior_transform: function
priot transorm function
ndim: int
number of dimension
"""
print("Run Nested Fitting for the image... #a of params: %d"%ndim)
start = time.time()
if n_cpu is None:
n_cpu = mp.cpu_count()-1
with mp.Pool(processes=n_cpu) as pool:
print("Opening pool: # of CPU used: %d"%(n_cpu))
pool.size = n_cpu
dlogz = 1e-3 * (nlive_init - 1) + 0.01
pdsampler = dynesty.DynamicNestedSampler(loglikelihood, prior_transform, ndim,
sample=sample, pool=pool,
use_pool={'update_bound': False})
pdsampler.run_nested(nlive_init=nlive_init,
nlive_batch=nlive_batch,
maxbatch=maxbatch,
print_progress=print_progress,
dlogz_init=dlogz,
wt_kwargs={'pfrac': pfrac})
end = time.time()
print("Finish Fitting! Total time elapsed: %.3gs"%(end-start))
return pdsampler
def __init__(self,
container,
sample='auto',
bound='multi',
n_cpu=None,
n_thread=None,
run=True,
results=None):
""" A class for runnning the sampling and plotting results """
# False if a previous run is read
self.run = run
self.container = container
self.image = container.image
self.ndim = container.ndim
self.labels = container.labels
if run:
if n_cpu is None:
n_cpu = min(mp.cpu_count() - 1, 10)
if n_thread is not None:
n_thread = max(n_thread, n_cpu - 1)
if n_cpu > 1:
self.open_pool(n_cpu)
self.use_pool = {'update_bound': False}
else:
self.pool = None
self.use_pool = None
self.prior_tf = container.prior_transform
self.loglike = container.loglikelihood
dsampler = dynesty.DynamicNestedSampler(self.loglike,
self.prior_tf,
self.ndim,
sample=sample,
bound=bound,
pool=self.pool,
queue_size=n_thread,
use_pool=self.use_pool)
self.dsampler = dsampler
else:
self._results = results # use existed results
def test_maxcall():
# test of maxcall functionality
ndim = 2
rstate = get_rstate()
sampler = dynesty.NestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
rstate=rstate)
sampler.run_nested(dlogz=1, maxcall=1000, print_progress=printing)
sampler = dynesty.DynamicNestedSampler(loglike,
prior_transform,
ndim,
nlive=nlive,
rstate=rstate)
sampler.run_nested(dlogz_init=1, maxcall=1000, print_progress=printing)
def test_inf():
# Test of logl that returns -inf
ndim = 2
rstate = get_rstate()
sampler = dynesty.NestedSampler(loglike_inf,
prior_transform,
ndim,
nlive=nlive,
rstate=rstate)
sampler.run_nested(print_progress=printing)
sampler = dynesty.DynamicNestedSampler(loglike_inf,
prior_transform,
ndim,
nlive=nlive,
rstate=rstate)
sampler.run_nested(dlogz_init=1, print_progress=printing)
def __init__(self, likelihood, parameters, **kwargs):
prop_default = {
"nlive": 1000,
"bound": "single",
"which_sampler": "dynamic",
"run_kwargs": {},
}
self.likelihood = likelihood
for prop, default in prop_default.items():
setattr(self, prop, kwargs.get(prop, default))
for prop, default in prop_default.items():
kwargs[prop] = kwargs.get(prop, default)
super().__init__(parameters, **kwargs)
current_point_dict = self.get_current_point()
self.current_point = np.array(
[current_point_dict[key] for key in self.key_order])
self.injection = self.current_point.copy()
if self.which_sampler == "dynamic":
print("Running dynamic sampler.")
self.sampler = dynesty.DynamicNestedSampler(
loglike,
ptform,
len(self.test_inds),
logl_args=(self.likelihood, ),
ptform_args=(self.sampling_values, self.key_order,
self.test_inds),
**kwargs)
elif self.which_sampler == "static":
print("Running static sampler.")
self.sampler = dynesty.NestedSampler(
loglike,
ptform,
len(self.test_inds),
logl_args=(self.likelihood, ),
ptform_args=(self.sampling_values, self.key_order,
self.test_inds),
**kwargs)
else:
raise ValueError("which_sampler must be dynamic or static.")
def test_usepool(func):
# test all the use_pool options, toggle them one by one
rstate = get_rstate()
use_pool = {}
for k in POOL_KW:
use_pool[k] = False
use_pool[func] = True
with mp.Pool(2) as pool:
sampler = dynesty.DynamicNestedSampler(loglike_gau,
prior_transform_gau,
ndim,
nlive=nlive,
rstate=rstate,
use_pool=use_pool,
pool=pool,
queue_size=100)
sampler.run_nested(maxiter=10000, print_progress=printing)