def main():
parser = MyParser()
parser.add_argument('config_fname',
help="full path to config filename",
nargs='?',
type=str)
parser.add_argument('redshift',
help="central redshift for plotting spectrum",
nargs='?',
type=float)
parser.add_argument('dv',
help="velocity range for spectrum (+/-dv)",
nargs='?',
type=float)
#parser.add_argument("--truths", nargs='+', type=float)
parser.add_argument('-t',
"--test",
help="plot data product of test",
action="store_true")
args = parser.parse_args()
if len(sys.argv) < 4 and not args.test:
parser.print_help(sys.stderr)
sys.exit(1)
if args.test:
from bayesvp.utilities import get_bayesvp_Dir
path = get_bayesvp_Dir()
config_fname = path + '/data/example/config_OVI.dat'
config_param = DefineParams(config_fname)
args.redshift = 0.0
args.dv = 300.0
else:
if os.path.isfile(args.config_fname):
config_param = DefineParams(args.config_fname)
else:
sys.exit('Config file does not exist:\n %s' % args.config_fname)
output_model = ProcessModel(config_param)
output_model.spline_binned_pdf()
output_model.plot_model_comparison(args.redshift, args.dv)
output_model.write_model_summary()
output_model.write_model_spectrum()
output_model.plot_gr_indicator()
#output_model.corner_plot(nbins=30,truths=args.truths)
output_model.corner_plot(nbins=30)
def setUp(self):
# read example config file
code_path = get_bayesvp_Dir()
self.config_ex = code_path + '/data/example/config_OVI.dat'
self.config_params = DefineParams(self.config_ex)
self.posterior = Posterior(self.config_params)
def bvp_process(config_fname, redshift, dv):
config_param = DefineParams(config_fname)
output_model = ProcessModel(config_param)
output_model.spline_binned_pdf()
output_model.plot_model_comparison(redshift, dv)
output_model.write_model_summary()
output_model.write_model_spectrum()
output_model.plot_gr_indicator()
output_model.corner_plot(nbins=30)
return None
# from bayesvp.vp_model import simple_spec
#-------------------------------------------------------------
# set the path of the data
spectrum_path = '/home/tanvir/BLA_project/codes/examples/'
# spectrum_path = 'E:/TANVIR_WORK/codes/examples/'
# create a config file
config_writer = wc.WriteBayesVPConfig().print_to_file("-i")
# load the config file
config_fname = spectrum_path + 'bvp_configs/config_HI.dat'
# extract all of the relevant information using the DefineParams function
config_params = DefineParams(config_fname)
#-------- run the MCMC fit --------------------------
# fitting log is written in 'config.log' file.
start = time.time()
#
mc_single(config_params)
mcmc(config_fname)
#
end = time.time()
print((end - start) / 60.)
#-----------------------------------------------------------
# The fitting process after the step is completed. The output chain is a binary file which ends with .npy
# contains all the information that we need.
# ---- read the .npy file ------
from bayesvp.mcmc_setup import bvp_mcmc as mcmc from bayesvp.config import DefineParams from bayesvp.scripts import bvp_process_model config_file = '/home/vikram/BLA_hub/test_vikram/hi_config.dat' # the mcmc fit mcmc(config_file, print_config=True) # after the chains x = DefineParams(config_file) op = bvp_process_model.ProcessModel(x) op.plot_model_comparison(0.081, 400) op.write_model_summary() op.write_model_spectrum() op.plot_gr_indicator() op.corner_plot()
def setUp(self):
code_path = get_bayesvp_Dir()
# read example config file
self.config_ex = code_path + '/data/example/config_OVI.dat'
self.config_params = DefineParams(self.config_ex)
def bvp_mcmc(config_fname, print_config=False):
"""
Run fixed number of component fit specified by the
config file or make copies of the configs and run up
to the maximum number of components until the best model
is found.
Parameters:
-----------
config_fname: str
Full path + the file name
See ./readme.md for detaieded structure of the config file
Returns
-----------
chains: python format (.npy) binary file
One or more chains for each MCMC run.
"""
# Determine if config is set to automatically choose the best model
# If so, create multiple configs file specified by n_component_max
auto_vp, n_component_min, n_component_max = determine_autovp(config_fname)
if auto_vp:
model_evidence = np.zeros(n_component_max - n_component_min + 1)
for n in xrange(n_component_max - n_component_min + 1):
# Get new config filename;
basename_with_path, config_extension = os.path.splitext(
config_fname)
config_fname_ncomp = (basename_with_path +
str(n + n_component_min) + config_extension)
# Load config parameter object
config_params = DefineParams(config_fname_ncomp)
# Run MCMC
bvp_mcmc_single(config_params, config_params.chain_fname)
# compute values of aic, bic or bf
model_evidence[n] = model_info_criterion(config_params)
components_count = np.arange(n_component_min, n_component_max + 1)
# compare with the previous fit
if 0 < n <= n_component_max - 1:
#if config_params.model_selection.lower() in ('odds','bf'):
# index = np.where(model_evidence[:n+1] == np.max(model_evidence[:n+1]))[0]
if config_params.model_selection.lower() in ('aic', 'bic'):
index = np.where(
model_evidence[:n + 1] == np.min(model_evidence[:n +
1]))[0]
# Compare BIC/AIC/Odds ratios
if compare_model(model_evidence[n - 1], model_evidence[n],
config_params.model_selection):
np.savetxt(config_params.mcmc_outputpath + '/' +
config_params.model_selection + '_' +
config_params.chain_short_fname[:-1] + '.dat',
np.c_[components_count[:n + 1],
model_evidence[:n + 1]],
fmt=('%d', '%.4f'),
header='nComponents\tValues')
break
else:
# Load config parameter object
config_params = DefineParams(config_fname)
if print_config:
config_params.print_config_params()
# Run fit as specified in config
bvp_mcmc_single(config_params)