diff --git a/.ci/bbtautau/klambda.json b/.ci/bbtautau/klambda.json
index c81d2e2d4141d62b15fe9618bf5ba46c36889694..7d94124ef835b838eace82109f416306dbc01fb4 100644
--- a/.ci/bbtautau/klambda.json
+++ b/.ci/bbtautau/klambda.json
@@ -1,27 +1,27 @@
{
- "nll": [
- 290961.64047893154,
- 290961.7870285931,
- 290961.65236822236
- ],
- "qmu": [
- 0.0,
- 0.2930993231711909,
- 0.02377858164254576
- ],
- "status": [
- 0,
- 0,
- 0
- ],
- "klambda": [
- 1.497080183188742,
- 0,
- 1
- ],
- "mu": [
- 1.497080183188742,
- 0,
- 1
- ]
+ "nll": [
+ 290961.64047893154,
+ 290961.7870285931,
+ 290961.65236822236
+ ],
+ "qmu": [
+ 0.0,
+ 0.2930993231711909,
+ 0.02377858164254576
+ ],
+ "status": [
+ 0,
+ 0,
+ 0
+ ],
+ "klambda": [
+ 1.497080183188742,
+ 0,
+ 1
+ ],
+ "mu": [
+ 1.497080183188742,
+ 0,
+ 1
+ ]
}
\ No newline at end of file
diff --git a/.ci/bbyy/klambda.json b/.ci/bbyy/klambda.json
index 7bc3217f3080c2bdd9d7cedca4d1272ffcbe0053..112fe12cf5654db3d39ec2a1e360ed85297751f1 100644
--- a/.ci/bbyy/klambda.json
+++ b/.ci/bbyy/klambda.json
@@ -1,27 +1,27 @@
{
- "nll": [
- 517.9102262602762,
- 518.7035184306969,
- 518.2302598267265
- ],
- "qmu": [
- 0.0,
- 1.5865843408414548,
- 0.64006713290064
- ],
- "status": [
- 0,
- 0,
- 0
- ],
- "klambda": [
- 2.7511436422832256,
- 0,
- 1
- ],
- "mu": [
- 2.7511436422832256,
- 0,
- 1
- ]
+ "nll": [
+ 517.9102262602762,
+ 518.7035184306969,
+ 518.2302598267265
+ ],
+ "qmu": [
+ 0.0,
+ 1.5865843408414548,
+ 0.64006713290064
+ ],
+ "status": [
+ 0,
+ 0,
+ 0
+ ],
+ "klambda": [
+ 2.7511436422832256,
+ 0,
+ 1
+ ],
+ "mu": [
+ 2.7511436422832256,
+ 0,
+ 1
+ ]
}
\ No newline at end of file
diff --git a/bin/quickstats b/bin/quickstats
index cb8479db8c7e480f16c8efe85090b25573067131..c7867dbe4c5a21931235fade30b78c9d9654705d 100755
--- a/bin/quickstats
+++ b/bin/quickstats
@@ -1,44 +1,6 @@
#! /usr/bin/env python
-import click
-from quickstats.clis.core import (run_pulls, plot_pulls, compile_macros,
- harmonize_np, cls_limit, generate_asimov,
- generate_standard_asimov, toy_significance,
- toy_limit, add_macro, remove_macro)
-from quickstats.clis.inspect_ws import inspect_ws
-from quickstats.clis.inspect_rfile import inspect_rfile
-from quickstats.clis.limit_setting import limit_scan
-from quickstats.clis.processor_cli import process_rfile
-from quickstats.clis.likelihood_scan import likelihood_scan
-from quickstats.clis.likelihood_fit import likelihood_fit, np_correlation
-from quickstats.clis.workspace_tools import build_xml_ws, modify_ws, compare_ws, combine_ws, decompose_ws
-
-@click.group()
-def cli():
- pass
+from quickstats.clis import cli
if __name__ == "__main__":
- cli.add_command(compile_macros)
- cli.add_command(add_macro)
- cli.add_command(remove_macro)
- cli.add_command(inspect_ws)
- cli.add_command(toy_significance)
- cli.add_command(toy_limit)
- cli.add_command(run_pulls)
- cli.add_command(plot_pulls)
- cli.add_command(likelihood_fit)
- cli.add_command(likelihood_scan)
- cli.add_command(cls_limit)
- cli.add_command(limit_scan)
- cli.add_command(np_correlation)
- cli.add_command(harmonize_np)
- cli.add_command(generate_asimov)
- cli.add_command(generate_standard_asimov)
- cli.add_command(inspect_rfile)
- cli.add_command(process_rfile)
- cli.add_command(build_xml_ws)
- cli.add_command(modify_ws)
- cli.add_command(compare_ws)
- cli.add_command(combine_ws)
- cli.add_command(decompose_ws)
cli()
\ No newline at end of file
diff --git a/quickstats/analysis/analysis_path_manager.py b/quickstats/analysis/analysis_path_manager.py
index a3dfef7f9304dad4a9fc5d6f3e2a432007503559..911a59e74eb422219efe6c846c21f36591d97b1a 100644
--- a/quickstats/analysis/analysis_path_manager.py
+++ b/quickstats/analysis/analysis_path_manager.py
@@ -58,24 +58,21 @@ class AnalysisPathManager(PathManager):
self.study_name = study_name
self.raw_base_path = base_path
self.update()
+
+ def get_base_path(self):
+ if self.study_name is None:
+ return self.base_path
+ if self.base_path is None:
+ return self.study_name
+ return os.path.join(self.base_name, self.study_name)
def update(self):
- if self.raw_base_path is None:
- if self.study_name is None:
- base_path = None
- else:
- base_path = self.study_name
- else:
- if self.study_name is None:
- base_path = self.raw_base_path
- else:
- base_path = os.path.join(self.raw_base_path, self.study_name)
- self.base_path = base_path
+ pass
def set_study_name(self, study_name:str):
self.study_name = study_name
self.update()
def set_base_path(self, base_path:str):
- self.raw_base_path = base_path
+ self.base_path = base_path
self.update()
\ No newline at end of file
diff --git a/quickstats/analysis/event_categorization.py b/quickstats/analysis/event_categorization.py
index b8e31b1b2ef88431b50824b8e0ed41033da5ea30..763e93e5bafcaf6f53b30465655c7ed15481356b 100644
--- a/quickstats/analysis/event_categorization.py
+++ b/quickstats/analysis/event_categorization.py
@@ -1094,12 +1094,12 @@ class EventCategorization(DataLoader):
for sample in samples:
category_df = self.get_category_df(sample, category_index)
key = sample
- if do_blind:
+ if self.do_blind:
key = self.get_blind_sample_name(sample)
if resampling:
category_df = self.get_resampled_df(category_df, random_state=resampling_random_state)
yields[category_name][key] = float(category_df[weight_name].sum())
- if is_data:
+ if self.is_data_sample(sample):
#calculate poisson interval
poisson_interval = get_poisson_interval([yields[category_name][key]])
yields_err[category_name][key] = {"errlo": poisson_interval["lo"][0],
diff --git a/quickstats/clis/__init__.py b/quickstats/clis/__init__.py
index e69de29bb2d1d6434b8b29ae775ad8c2e48c5391..f4b14a71589da4d55ad71473bee7743beaf284eb 100644
--- a/quickstats/clis/__init__.py
+++ b/quickstats/clis/__init__.py
@@ -0,0 +1,8 @@
+from .core import *
+from .workspace_tools import *
+from .limit_setting import *
+from .likelihood_tools import *
+from .nuisance_parameter_tools import *
+from .stat_tools import *
+from .inspect_rfile import *
+from .processor_tools import *
\ No newline at end of file
diff --git a/quickstats/clis/core.py b/quickstats/clis/core.py
index 7854455e63a46efd3fc950ee0e38b559d06b2889..1eba744db508f37dc37354f0790b7e4aba95cb1a 100644
--- a/quickstats/clis/core.py
+++ b/quickstats/clis/core.py
@@ -2,6 +2,30 @@ import os
import json
import click
+__all__ = ['compile_macros', 'add_macro', 'remove_macro', 'cli']
+
+class NaturalOrderGroup(click.Group):
+ """Command group trying to list subcommands in the order they were added.
+
+ Make sure you initialize the `self.commands` with OrderedDict instance.
+
+ With decorator, use::
+
+ @click.group(cls=NaturalOrderGroup, commands=OrderedDict())
+ """
+
+ def list_commands(self, ctx):
+ """List command names as they are in commands dict.
+
+ If the dict is OrderedDict, it will preserve the order commands
+ were added.
+ """
+ return self.commands.keys()
+
+@click.group(cls=NaturalOrderGroup)
+def cli():
+ pass
+
class DelimitedStr(click.Option):
def type_cast_value(self, ctx, value):
try:
@@ -9,281 +33,8 @@ class DelimitedStr(click.Option):
except:
raise click.BadParameter(value)
-@click.command(name='run_pulls')
-@click.option('-i', '--input_file', 'filename', required=True,
- help='Path to the input workspace file')
-@click.option('-x', '--poi', 'poi_name', default=None,
- help='POI to measure NP impact on')
-@click.option('-o', '--outdir', default="pulls", show_default=True,
- help='Output directory')
-@click.option('-w', '--workspace', 'ws_name', default=None,
- help='Name of workspace. Auto-detect by default.')
-@click.option('-m', '--model_config', 'mc_name', default=None,
- help='Name of model config. Auto-detect by default.')
-@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
- help='Name of dataset')
-@click.option('--filter', 'filter_expr', default=None, show_default=True,
- help='Filter nuisance parameter(s) to run pulls and impacts on.'+\
- 'Multiple parameters are separated by commas.'+\
- 'Wildcards are accepted. All NPs are included by default.')
-@click.option('-r', '--profile', 'profile_param', default=None, show_default=True,
- help='Parameters to profile')
-@click.option('-f', '--fix', 'fix_param', default=None, show_default=True,
- help='Parameters to fix')
-@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
- help='Name of initial snapshot')
-@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
- help='Minimizer type')
-@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
- help='Minimizer algorithm')
-@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
- help='Number of CPUs to use per parameter')
-@click.option('--binned/--unbinned', 'binned_likelihood', default=True, show_default=True,
- help='Binned likelihood')
-@click.option('-q', '--precision', type=float, default=0.001, show_default=True,
- help='Precision for scan')
-@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
- help='Convergence criterium')
-@click.option('--retry', type=int, default=2, show_default=True,
- help='Maximum number of retries upon a failed fit')
-@click.option('--print_level', type=int, default=-1, show_default=True,
- help='Minimizer print level')
-@click.option('--strategy', type=int, default=1, show_default=True,
- help='Default strategy')
-@click.option('--fix-cache/--no-fix-cache', default=True, show_default=True,
- help='Fix StarMomentMorph cache')
-@click.option('--fix-multi/--no-fix-multi', default=True, show_default=True,
- help='Fix MultiPdf level 2')
-@click.option('--offset/--no-offset', default=True, show_default=True,
- help='Offset likelihood')
-@click.option('--optimize', type=int, default=2, show_default=True,
- help='Optimize constant terms')
-@click.option('--max_calls', type=int, default=-1, show_default=True,
- help='Maximum number of function calls')
-@click.option('--max_iters', type=int, default=-1, show_default=True,
- help='Maximum number of Minuit iterations')
-@click.option('--batch_mode/--no-batch', default=False, show_default=True,
- help='Batch mode when evaluating likelihood')
-@click.option('--int_bin_precision', type=float, default=-1., show_default=True,
- help='Integrate the PDF over the bins instead of using the probability '
- 'density at the bin centre')
-@click.option('--parallel', type=int, default=-1, show_default=True,
- help='\b\n Parallelize job across the N workers.'
- '\b\n Case 0: Jobs are run sequentially (for debugging).'
- '\b\n Case -1: Jobs are run across N_CPU workers.')
-@click.option('--cache/--no-cache', default=True, show_default=True,
- help='Cache existing result')
-@click.option('--exclude', 'exclude_expr', default=None, show_default=True,
- help='Exclude NPs to run pulls and impacts on. '+\
- 'Multiple parameters are separated by commas.'+\
- 'Wildcards are accepted.')
-@click.option('--save_log/--skip_log', default=True, show_default=True,
- help='Save log file.')
-@click.option('--constrained_only/--any_nuis', default=True, show_default=True,
- help='Whether to include constrained nuisance parameters only')
-@click.option('--version', type=click.Choice(['1', '2']), default='2', show_default=True,
- help='Version of tool to use (Choose between 1 and 2).')
-@click.option('-v', '--verbosity', default='INFO', show_default=True,
- type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
- help='Verbosity level.')
-def run_pulls(**kwargs):
- """
- Tool for computing NP pulls and impacts
- """
- version = kwargs.pop('version')
- if version == '1':
- from quickstats.components import NuisanceParameterPull
- NuisanceParameterPull().run_pulls(**kwargs)
- elif version == '2':
- from quickstats.concurrent import NuisanceParameterRankingRunner
- init_kwargs = {}
- for key in ['filename', 'filter_expr', 'exclude_expr', 'poi_name',
- 'data_name', 'cache', 'outdir', 'constrained_only',
- 'save_log', 'parallel', 'verbosity']:
- init_kwargs[key] = kwargs.pop(key)
- init_kwargs['config'] = kwargs
- runner = NuisanceParameterRankingRunner(**init_kwargs)
- runner.run()
-
-@click.command(name='plot_pulls')
-@click.option('-i', '--inputdir', required=True, help='Path to directory containing pull results')
-@click.option('-p', '--poi', default=None, help='Parameter of interest for plotting impact')
-@click.option('-n', '--n_rank', type=int, default=None, help='Total number of NP to rank')
-@click.option('-m', '--rank_per_plot', type=int, default=20, show_default=True,
- help='Number of NP to show in a single plot')
-@click.option('--ranking/--no_ranking', default=True, show_default=True,
- help='Rank NP by impact')
-@click.option('--threshold', type=float, default=0., show_default=True,
- help='Filter NP by postfit impact threshold')
-@click.option('--show_sigma/--hide_sigma', default=True, show_default=True,
- help='Show one standard deviation pull')
-@click.option('--show_prefit/--hide_prefit', default=True, show_default=True,
- help='Show prefit impact')
-@click.option('--show_postfit/--hide_postfit', default=True, show_default=True,
- help='Show postfit impact')
-@click.option('--sigma_bands/--no_sigma_bands', default=False, show_default=True,
- help='Draw +-1, +-2 sigma bands')
-@click.option('--sigma_lines/--no_sigma_lines', default=True, show_default=True,
- help='Draw +-1 sigma lines')
-@click.option('--ranking_label/--no_ranking_label', default=True, show_default=True,
- help='Show ranking labels')
-@click.option('--shade/--no_shade', default=True, show_default=True,
- help='Draw shade')
-@click.option('--correlation/--no_correlation', default=True, show_default=True,
- help='Show correlation impact')
-@click.option('--onesided/--overlap', default=True, show_default=True,
- help='Show onesided impact')
-@click.option('--relative/--absolute', default=False, show_default=True,
- help='Show relative variation')
-@click.option('--theta_max', type=float, default=2, show_default=True,
- help='Pull range')
-@click.option('-y', '--padding', type=int, default=7, show_default=True,
- help='Padding below plot for texts and legends. NP column height is 1 unit.')
-@click.option('-h', '--height', type=float, default=1.0, show_default=True,
- help='NP column height')
-@click.option('-s', '--spacing', type=float, default=0., show_default=True,
- help='Spacing between impact box')
-@click.option('--label-fontsize', type=float, default=20., show_default=True,
- help='Fontsize of analysis label text')
-@click.option('-d', '--display_poi', default=r"$\mu$", show_default=True,
- help='POI name to be shown in the plot')
-@click.option('-t', '--extra_text', default=None, help='Extra texts below the ATLAS label. '+\
- 'Use "//" as newline delimiter')
-@click.option('--elumi_label/--no_elumi_label', default=True, show_default=True,
- help='Show energy and luminosity labels')
-@click.option('--ranking_label/--no_ranking_label', default=True, show_default=True,
- help='Show ranking label')
-@click.option('--energy', default="13 TeV", show_default=True,
- help='Beam energy')
-@click.option('--lumi', default="140 fb$^{-1}$", show_default=True,
- help='Luminosity')
-@click.option('--status', default="int", show_default=True,
- help='\b\n Analysis status. Choose from'
- '\b\n int : Internal'
- '\b\n wip : Work in Progress'
- '\b\n prelim : Preliminary'
- '\b\n final : *no status label*'
- '\b\n *custom input* : *custom input*')
-@click.option('--combine_pdf/--split_pdf', default=True, show_default=True,
- help='Combine all ranking plots into a single pdf')
-@click.option('--outdir', default='ranking_plots', show_default=True,
- help='Output directory')
-@click.option('-o', '--outname', default='ranking', show_default=True,
- help='Output file name prefix')
-@click.option('--style', default='default', show_default=True,
- help='Plotting style. Built-in styles are "default" and "trex".'+\
- 'Specify path to yaml file to set custom plotting style.')
-@click.option('--fix_axis_scale/--free_axis_scale', default=True, show_default=True,
- help='Fix the axis scale across all ranking plots')
-@click.option('--version', type=click.Choice(['1', '2']), default='2', show_default=True,
- help='Version of tool to use (Choose between 1 and 2).')
-def plot_pulls(**kwargs):
- """
- Tool for plotting NP pulls and impact rankings
- """
- from quickstats.plots.np_ranking_plot import NPRankingPlot
- inputdir, poi = kwargs.pop('inputdir'), kwargs.pop('poi')
- version = kwargs.pop('version')
- ranking_plot = NPRankingPlot(inputdir, poi, version=version)
- ranking_plot.plot(**kwargs)
-
-@click.command(name='cls_limit')
-@click.option('-i', '--input_file', 'filename', required=True,
- help='Path to the input workspace file')
-@click.option('-p', '--poi', 'poi_name', default=None, show_default=True,
- help='POI to scan. If not specified, the first POI from the workspace is used.')
-@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
- help='Name of dataset')
-@click.option('--asimov_data_name', 'asimov_data_name', default=None,
- help='If given, use custom background asimov dataset instead of generating on the fly.')
-@click.option('-o', '--outname', default='limits.json', show_default=True,
- help='Name of output')
-@click.option('--mu_exp', type=float, default=0, show_default=True,
- help='Expected signal strengh value to be used for Asimov generation')
-@click.option('--blind/--unblind', 'do_blind', default=True, show_default=True,
- help='Blind/unblind analysis')
-@click.option('--CL', 'CL', type=float, default=0.95, show_default=True,
- help='CL value to use')
-@click.option('--precision', default=0.005, show_default=True,
- help='precision in mu that defines iterative cutoff')
-@click.option('--adjust_fit_range/--keep_fit_range', default=True, show_default=True,
- help='whether to adjust the fit range to median limit +- 5 sigma for observed fit')
-@click.option('--do_tilde/--no_tilde', default=True, show_default=True,
- help='bound mu at zero if true and do the \tilde{q}_{mu} asymptotics')
-@click.option('--predictive_fit/--no_predictive_fit', default=False, show_default=True,
- help='extrapolate best fit nuisance parameters based on previous fit results')
-@click.option('--do_better_bands/--skip_better_bands', default=True, show_default=True,
- help='evaluate asymptotic CLs limit for various sigma bands')
-@click.option('--better_negative_bands/--skip_better_negative_bands', default=False, show_default=True,
- help='evaluate asymptotic CLs limit for negative sigma bands')
-@click.option('--binned/--unbinned', 'binned_likelihood', default=True, show_default=True,
- help='Binned likelihood')
-@click.option('--save_summary/--skip_summary', default=True, show_default=True,
- help='Save summary information')
-@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
- help='Parameters to fix')
-@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
- help='Parameters to profile')
-@click.option('-w', '--workspace', 'ws_name', default=None, show_default=True,
- help='Name of workspace. Auto-detect by default.')
-@click.option('-m', '--model_config', 'mc_name', default=None, show_default=True,
- help='Name of model config. Auto-detect by default.')
-@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
- help='Name of initial snapshot')
-@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
- help='Minimizer type')
-@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
- help='Minimizer algorithm')
-@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
- help='Convergence criterium')
-@click.option('--retry', type=int, default=2, show_default=True,
- help='Maximum number of retries upon a failed fit')
-@click.option('--strategy', type=int, default=1, show_default=True,
- help='Default minimization strategy')
-@click.option('--print_level', type=int, default=-1, show_default=True,
- help='Minimizer print level')
-@click.option('--timer/--no_timer', default=False, show_default=True,
- help='Enable minimizer timer')
-@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
- help='Number of CPUs to use per parameter')
-@click.option('--offset/--no-offset', default=True, show_default=True,
- help='Offset likelihood')
-@click.option('--optimize', type=int, default=2, show_default=True,
- help='Optimize constant terms')
-@click.option('--improve', type=int, default=0, show_default=True,
- help='Execute improve after each minimization')
-@click.option('--minimizer_offset', type=int, default=1, show_default=True,
- help='Enable minimizer offsetting')
-@click.option('--fix-cache/--no-fix-cache', default=True, show_default=True,
- help='Fix StarMomentMorph cache')
-@click.option('--fix-multi/--no-fix-cache', default=True, show_default=True,
- help='Fix MultiPdf level 2')
-@click.option('--max_calls', type=int, default=-1, show_default=True,
- help='Maximum number of function calls')
-@click.option('--max_iters', type=int, default=-1, show_default=True,
- help='Maximum number of Minuit iterations')
-@click.option('--batch_mode/--no-batch', default=False, show_default=True,
- help='Batch mode when evaluating likelihood')
-@click.option('--int_bin_precision', type=float, default=-1., show_default=True,
- help='Integrate the PDF over the bins instead of using the probability '
- 'density at the bin centre')
-@click.option('--constrain/--no-constrain', 'constrain_nuis', default=True, show_default=True,
- help='Use constrained NLL')
-@click.option('-v', '--verbosity', default='INFO', show_default=True,
- type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
- help='Verbosity level.')
-def cls_limit(**kwargs):
- """
- Tool for evaluating Asymptotic CLs limit
- """
- from quickstats.components import AsymptoticCLs
- outname = kwargs.pop('outname')
- save_summary = kwargs.pop('save_summary')
- asymptotic_cls = AsymptoticCLs(**kwargs)
- asymptotic_cls.evaluate_limits()
- asymptotic_cls.save(outname, summary=save_summary)
-@click.command(name='compile')
+@cli.command(name='compile')
@click.option('-m', '--macros', default=None, show_default=True,
help='Macros to compile (separated by commas). By default all macros are compiled.')
def compile_macros(macros):
@@ -293,14 +44,14 @@ def compile_macros(macros):
import quickstats
quickstats.compile_macros(macros)
-@click.command(name='add_macro')
+@cli.command(name='add_macro')
@click.option('-i', '--input_path', 'path', required=True,
help='Path to the directory containing the source file for the macro.')
-@click.option('-n', '--name',
+@click.option('-n', '--name', default=None,
help='Name of the macro. By default, the name of the input directory is used.')
@click.option('-f', '--force', is_flag=True,
help='Force overwrite existing files.')
-@click.option('--copy-files/--do-not-copy-files', 'copy_files', default=True,
+@click.option('--copy-files/--do-not-copy-files', 'copy_files', default=True, show_default=True,
help='Whether to copy files from the input directory (required if not already copied).')
@click.option('--add-to-workspace-extension/--do-not-add-to-workspace-extension', 'workspace_extension',
default=True, show_default=True,
@@ -312,7 +63,7 @@ def add_macro(**kwargs):
import quickstats
quickstats.add_macro(**kwargs)
-@click.command(name='remove_macro')
+@cli.command(name='remove_macro')
@click.option('-n', '--name', required=True,
help='Name of the macro.')
@click.option('-f', '--force', is_flag=True,
@@ -325,298 +76,4 @@ def remove_macro(**kwargs):
Remove a ROOT macro from the module
"""
import quickstats
- quickstats.remove_macro(**kwargs)
-
-@click.command(name='harmonize_np')
-@click.argument('ws_files', nargs=-1)
-@click.option('-r', '--reference', required=True, help='Path to reference json file containing renaming scheme')
-@click.option('-i', '--input_config_path', default=None, show_default=True,
- help='Path to json file containing input workspace paths')
-@click.option('-b', '--base_path', default='./', show_default=True,
- help='Base path for input config')
-@click.option('-o', '--outfile', default='renamed_np.json', show_default=True,
- help='Output filename')
-def harmonize_np(ws_files, reference, input_config_path, base_path, outfile):
- """
- Harmonize NP names across different workspaces
- """
- from quickstats.components import NuisanceParameterHarmonizer
- harmonizer = NuisanceParameterHarmonizer(reference)
- if (len(ws_files) > 0) and input_config_path is not None:
- raise RuntimeError('either workspace paths or json file containing workspace paths should be given')
- if len(ws_files) > 0:
- harmonizer.harmonize(ws_files, outfile=outfile)
- elif (input_config_path is not None):
- harmonizer.harmonize_multi_input(input_config_path, base_path, outfile=outfile)
-
-
-@click.command(name='generate_asimov')
-@click.option('-i', '--input_file', 'filename', required=True,
- help='Path to the input workspace file.')
-@click.option('-o', '--output_file', 'outname', required=True,
- help='Name of the output workspace containing the '
- 'generated asimov dataset.')
-@click.option('-p', '--poi', required=True,
- help='Name of the parameter of interest (POI).')
-@click.option('--poi_val', type=float, default=None, show_default=True,
- help='Generate asimov data with POI set at the specified value. '
- 'If None, POI will be kept at the post-fit value if a fitting '
- 'is performed or the pre-fit value if no fitting is performed.')
-@click.option('--poi_profile', type=float, default=None, show_default=True,
- help='Perform nuisance parameter profiling with POI set at the specified value. '
- 'This option is only effective if do_fit is set to True. If None, POI is '
- 'set floating (i.e. unconditional maximum likelihood estimate).')
-@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
- help='Parameters to fix')
-@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
- help='Parameters to profile')
-@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
- help='Name of initial snapshot')
-@click.option('--modify-globs/--keep-globs', default=True, show_default=True,
- help='Match the values of nuisance parameters and the corresponding global '
- 'observables when generating the asimov data. This is important for making '
- 'sure the asimov data has the (conditional) minimal NLL.')
-@click.option('--do-fit/--no-fit', default=True, show_default=True,
- help='Perform nuisance parameter profiling with a fit to the given dataset.')
-@click.option('--asimov_name', default="asimovData_{mu}", show_default=True,
- help='Name of the generated asimov dataset.')
-@click.option('--asimov_snapshot', default="asimovData_{mu}", show_default=True,
- help='Name of the snapshot that generates the asimov dataset.')
-@click.option('-d', '--data', default='combData', show_default=True,
- help='Name of the dataset used in NP profiling.')
-@click.option('--constraint_option', default=0, show_default=True,
- help='\b\n Customize the target of nuisance paramaters involved in the profiling.'
- '\b\n Case 0: All nuisance parameters are allowed to float;'
- '\b\n Case 1: Constrained nuisance parameters are fixed to 0.'
- '\b\n Unconstrained nuisrance parameters are allowed to float.')
-@click.option('-c', '--configuration', default=None,
- help='Path to the json configuration file containing'
- ' the minimizer options for NP profiling.')
-@click.option('--rebuild/--do-not-rebuild', default=False, show_default=True,
- help='Rebuild the workspace.')
-@click.option('-v', '--verbosity', default='INFO', show_default=True,
- type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
- help='Verbosity level.')
-def generate_asimov(**kwargs):
- """
- Generate Asimov dataset
- """
- from quickstats.components import AnalysisBase
- filename = kwargs.pop('filename')
- data_name = kwargs.pop('data')
- poi_name = kwargs.pop('poi')
- outname = kwargs.pop('outname')
- verbosity = kwargs.pop('verbosity')
- fix_param = kwargs.pop('fix_param')
- profile_param = kwargs.pop('profile_param')
- snapshot_name = kwargs.pop('snapshot_name')
- config_file = kwargs.pop('configuration')
- rebuild = kwargs.pop('rebuild')
- if config_file is not None:
- config = json.load(open(config_file, 'r'))
- else:
- config = {}
- config['fix_param'] = fix_param
- config['profile_param'] = profile_param
- config['snapshot_name'] = snapshot_name
- asimov_config = {
- "poi_val" : kwargs.pop("poi_val"),
- "poi_profile" : kwargs.pop("poi_profile"),
- "do_fit" : kwargs.pop("do_fit"),
- "modify_globs" : kwargs.pop("modify_globs"),
- "do_import" : True,
- "asimov_name" : kwargs.pop("asimov_name"),
- "asimov_snapshot" : kwargs.pop("asimov_snapshot"),
- "constraint_option" : kwargs.pop("constraint_option"),
- "restore_states" : 0
- }
- analysis = AnalysisBase(filename, poi_name=poi_name, data_name=data_name,
- config=config, verbosity=verbosity)
- analysis.generate_asimov(**asimov_config)
- analysis.save(outname, rebuild=rebuild)
-
-@click.command(name='generate_standard_asimov')
-@click.option('-i', '--input_file', 'filename', required=True,
- help='Path to the input workspace file.')
-@click.option('-o', '--output_file', 'outname', required=True,
- help='Name of the output workspace containing the '
- 'generated asimov dataset.')
-@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
- help='Name of the dataset used in NP profiling.')
-@click.option('-p', '--poi', 'poi_name', required=True,
- help='Name of the parameter of interest (POI). Multiple POIs are separated by commas.')
-@click.option('-s', '--poi_scale', type=float, default=1.0, show_default=True,
- help='Scale factor applied to the poi value')
-@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
- help='Parameters to fix')
-@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
- help='Parameters to profile')
-@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
- help='Name of initial snapshot')
-@click.option('--rebuild/--do-not-rebuild', default=False, show_default=True,
- help='Rebuild the workspace.')
-@click.option('--asimov_names', default=None, show_default=True,
- help='Names of the output asimov datasets (separated by commas). If not specified, '
- 'a default name for the corresponding asimov type will be given.')
-@click.option('--asimov_snapshots', default=None, show_default=True,
- help='Names of the output asimov snapshots (separated by commas). If not specified, '
- 'a default name for the corresponding asimov type will be given.')
-@click.option('-t', '--asimov_types', default="0,1,2", show_default=True,
- help='\b\n Types of asimov dataset to generate separated by commas.'
- '\b\n 0: fit with POI fixed to 0'
- '\b\n 1: fit with POI fixed to 1'
- '\b\n 2: fit with POI free and set POI to 1 after fit'
- '\b\n 3: fit with POI and constrained NP fixed to 0'
- '\b\n 4: fit with POI fixed to 1 and constrained NP fixed to 0'
- '\b\n 5: fit with POI free and constrained NP fixed to 0 and set POI to 1 after fit'
- '\b\n -1: nominal NP with POI set to 0'
- '\b\n -2: nominal NP with POI set to 1')
-@click.option('-v', '--verbosity', default='INFO', show_default=True,
- type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
- help='Verbosity level.')
-def generate_standard_asimov(**kwargs):
- """
- Generate standard Asimov dataset
- """
- from quickstats.components import AsimovGenerator
- from quickstats.utils.string_utils import split_str
- outname = kwargs.pop('outname')
- asimov_types = kwargs.pop('asimov_types')
- try:
- asimov_types = split_str(asimov_types, sep=",", cast=int)
- except:
- asimov_types = split_str(asimov_types, sep=",")
- fix_param = kwargs.pop('fix_param')
- profile_param = kwargs.pop('profile_param')
- snapshot_name = kwargs.pop('snapshot_name')
- poi_scale = kwargs.pop("poi_scale")
- asimov_names = kwargs.pop("asimov_names")
- asimov_snapshots = kwargs.pop("asimov_snapshots")
- verbosity = kwargs.pop("verbosity")
- rebuild = kwargs.pop("rebuild")
- kwargs['poi_name'] = split_str(kwargs.pop('poi_name'), sep=",")
- config = {
- 'fix_param': fix_param,
- 'profile_param': profile_param,
- 'snapshot_name': snapshot_name
- }
- from quickstats.utils.string_utils import split_str
- if asimov_names is not None:
- asimov_names = split_str(asimov_names, sep=",")
- if asimov_snapshots is not None:
- asimov_snapshots = split_str(asimov_snapshots, sep=",")
- generator = AsimovGenerator(**kwargs, config=config, verbosity=verbosity)
- generator.generate_standard_asimov(asimov_types, poi_scale=poi_scale,
- asimov_names=asimov_names,
- asimov_snapshots=asimov_snapshots)
- generator.save(outname, rebuild=rebuild)
-
-@click.command(name='toy_significance')
-@click.option('-i', '--input_file', 'filename', required=True,
- help='Path to the input workspace file.')
-@click.option('-o', '--output_file', 'outname', default="toy_study/results.json",
- help='Name of the output file containing toy results.')
-@click.option('-n', '--n_toys', type=int,
- help='Number of the toys to use.')
-@click.option('-b', '--batchsize', type=int, default=100, show_default=True,
- help='Divide the task into batches each containing this number of toys. '
- 'Result from each batch is saved for caching and different batches '
- 'are run in parallel if needed')
-@click.option('-s', '--seed', type=int, default=0, show_default=True,
- help='Random seed used for generating toy datasets.')
-@click.option('-p', '--poi', 'poi_name', default=None,
- help='Name of the parameter of interest (POI). If None, the first POI is used.')
-@click.option('-v', '--poi_val', type=float, default=0, show_default=True,
- help='POI value when generating the toy dataset.')
-@click.option('--binned/--unbinned', default=True, show_default=True,
- help='Generate binned toy dataset.')
-@click.option('--cache/--no-cache', default=True, show_default=True,
- help='Cache existing batch results.')
-@click.option('--fit_options', default=None, help='A json file specifying the fit options.')
-@click.option('-v', '--verbosity', default='INFO', show_default=True,
- type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
- help='Verbosity level.')
-@click.option('--parallel', type=int, default=-1, show_default=True,
- help='\b\n Parallelize job across the N workers.'
- '\b\n Case 0: Jobs are run sequentially (for debugging).'
- '\b\n Case -1: Jobs are run across N_CPU workers.')
-def toy_significance(**kwargs):
- """
- Generate toys and evaluate significance
- """
- from quickstats.components import PValueToys
- n_toys = kwargs.pop("n_toys")
- batchsize = kwargs.pop("batchsize")
- seed = kwargs.pop("seed")
- cache = kwargs.pop("cache")
- outname = kwargs.pop("outname")
- parallel = kwargs.pop("parallel")
- pvalue_toys = PValueToys(**kwargs)
- pvalue_toys.get_toy_results(n_toys=n_toys, batchsize=batchsize, seed=seed,
- cache=cache, save_as=outname, parallel=parallel)
-
-
-
-@click.command(name='toy_limit')
-@click.option('-i', '--input_file', 'filename', required=True,
- help='Path to the input workspace file.')
-@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
- help='Name of the dataset used for computing observed limit.')
-@click.option('-o', '--output_file', 'outname',
- default="toy_study/toy_result_seed_{seed}_batch_{batch}.root",
- show_default=True,
- help='Name of the output file containing toy results.')
-@click.option('--poi_max', type=float, default=None,
- help='Maximum range of POI.')
-@click.option('--poi_min', type=float, default=None,
- help='Minimum range of POI.')
-@click.option('--scan_max', type=float, default=None,
- help='Maximum scan value of POI.')
-@click.option('--scan_min', type=float, default=None,
- help='Minimum scan value of POI.')
-@click.option('--steps', type=int, default=10, show_default=True,
- help='Number of scan steps.')
-@click.option('--mu_val', type=float, default=None,
- help='Value of POI for running a single point')
-@click.option('-n', '--n_toys', type=int,
- help='Number of the toys to use.')
-@click.option('-b', '--batchsize', type=int, default=50, show_default=True,
- help='Divide the task into batches each containing this number of toys. '
- 'Result from each batch is saved for caching and different batches '
- 'are run in parallel if needed')
-@click.option('-s', '--seed', type=int, default=2021, show_default=True,
- help='Random seed used for generating toy datasets.')
-@click.option('-t', '--tolerance', type=float, default=1., show_default=True,
- help='Tolerance for minimization.')
-@click.option('-p', '--poi', 'poi_name', default=None,
- help='Name of the parameter of interest (POI). If None, the first POI is used.')
-@click.option('--minimizer_type', default="Minuit2", show_default=True,
- help='Minimizer type')
-@click.option('--strategy', type=int, default=1, show_default=True,
- help='Default minimization strategy')
-@click.option('--offset/--no-offset', default=True, show_default=True,
- help='Use NLL offset.')
-@click.option('--print_level', type=int, default=-1, show_default=True,
- help='Minimizer print level')
-@click.option('-v', '--verbosity', default='INFO', show_default=True,
- type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
- help='Verbosity level.')
-@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
- help='Parameters to fix')
-@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
- help='Parameters to profile')
-@click.option('--snapshot', 'snapshot_name', default=None, help='Name of initial snapshot')
-@click.option('--parallel', type=int, default=-1, show_default=True,
- help='\b\n Parallelize job across the N workers.'
- '\b\n Case 0: Jobs are run sequentially (for debugging).'
- '\b\n Case -1: Jobs are run across N_CPU workers.')
-def toy_limit(**kwargs):
- """
- Generate toys and evaluate limits
- """
- from quickstats.components.toy_limit_calculator import evaluate_batched_toy_limits
- if not (((kwargs['scan_min'] is None) and (kwargs['scan_max'] is None) and (kwargs['mu_val'] is not None)) or \
- ((kwargs['scan_min'] is not None) and (kwargs['scan_max'] is not None) and (kwargs['mu_val'] is None))):
- raise ValueError("please provide either (scan_min, scan_max, steps) for running a scan or (mu_val)"
- " for running a single point")
- evaluate_batched_toy_limits(**kwargs)
\ No newline at end of file
+ quickstats.remove_macro(**kwargs)
\ No newline at end of file
diff --git a/quickstats/clis/inspect_rfile.py b/quickstats/clis/inspect_rfile.py
index d2a5a93a259ad312cf5b43f90034b3244e12b0dc..81a7813f50ea854947dbefe4f77d5ea9888d76a2 100644
--- a/quickstats/clis/inspect_rfile.py
+++ b/quickstats/clis/inspect_rfile.py
@@ -1,8 +1,11 @@
import click
-
import fnmatch
+
+from .core import cli
+
+__all__ = ['inspect_rfile']
-@click.command(name='inspect_rfile')
+@cli.command(name='inspect_rfile')
@click.option('-i', '--file_expr', required=True, help='File name expression '
'(supports wild card, multiple files separated by commas).')
@click.option('-t', '--tree_name', required=True, help='Tree name.')
diff --git a/quickstats/clis/inspect_ws.py b/quickstats/clis/inspect_ws.py
deleted file mode 100644
index e977195f49bb07fc5bc4847a80972f75b11e3bde..0000000000000000000000000000000000000000
--- a/quickstats/clis/inspect_ws.py
+++ /dev/null
@@ -1,41 +0,0 @@
-import click
-
-import fnmatch
-
-from .core import DelimitedStr
-
-kItemChoices = ['workspace', 'dataset', 'snapshot', 'category', 'poi',
- 'detailed_nuisance_parameter', 'nuisance_parameter',
- 'global_observable', 'auxiliary']
-kDefaultItems = ",".join(['workspace', 'dataset', 'snapshot', 'category',
- 'poi', 'detailed_nuisance_parameter'])
-
-@click.command(name='inspect_ws')
-@click.option('-i', '--input_file', required=True, help='Path to the input workspace file')
-@click.option('-w', '--workspace', 'ws_name', default=None, help='Name of workspace. Auto-detect by default.')
-@click.option('-d', '--dataset', 'data_name', default=None, help='Name of dataset. Generally not needed.')
-@click.option('-m', '--model_config', 'mc_name', default=None, help='Name of model config. Auto-detect by default.')
-@click.option('-o', '--output_file', default=None, help='Export output to text file. If None, no output is saved.')
-@click.option('--items', cls=DelimitedStr, type=click.Choice(kItemChoices), show_default=True,
- default=kDefaultItems, help='Items to include in the summary (separated by commas).')
-@click.option('--include', 'include_patterns', default=None,
- help='Match variable names with given patterns (separated by commas).')
-@click.option('--exclude', 'exclude_patterns', default=None,
- help='Exclude variable names with given patterns (separated by commas).')
-@click.option('--detailed/--name-only', default=True, show_default=True,
- help='Include detailed variable properties or just the variable name in the summary.')
-def inspect_ws(input_file, ws_name=None, data_name=None, mc_name=None, output_file=None, items=None,
- include_patterns=None, exclude_patterns=None, detailed=True):
- '''
- Inspect workspace attributes
- '''
- from quickstats.components import ExtendedModel
- model = ExtendedModel(input_file, ws_name=ws_name, mc_name=mc_name, data_name=data_name,
- verbosity="WARNING")
- from quickstats.utils.string_utils import split_str
- #items = items.split(",") if items is not None else None
- include_patterns = split_str(include_patterns, ',') if include_patterns is not None else None
- exclude_patterns = split_str(exclude_patterns, ',') if exclude_patterns is not None else None
- model.stdout.verbosity = "INFO"
- model.print_summary(items=items, save_as=output_file, detailed=detailed,
- include_patterns=include_patterns, exclude_patterns=exclude_patterns)
\ No newline at end of file
diff --git a/quickstats/clis/likelihood_fit.py b/quickstats/clis/likelihood_fit.py
deleted file mode 100644
index ed3043d8d18b755f34e23632e84dc69c24536cbb..0000000000000000000000000000000000000000
--- a/quickstats/clis/likelihood_fit.py
+++ /dev/null
@@ -1,296 +0,0 @@
-import os
-import json
-import click
-
-@click.command(name='likelihood_fit')
-@click.option('-i', '--input_file', "filename", required=True,
- help='Path to the input workspace file.')
-@click.option('-o', '--outname', default='fit_result.json', show_default=True,
- help='Name of output file.')
-@click.option('--display/--no-display', default=True, show_default=True,
- help='Display fit result.')
-@click.option('--save/--no-save', "save_result", default=False, show_default=True,
- help='Save fit result.')
-@click.option('--save_log/--skip_log', default=False, show_default=True,
- help='Save log file.')
-@click.option('--save_ws', default=None, show_default=True,
- help='Save fitted workspace to a given path.')
-@click.option('--save_snapshot', default=None, show_default=True,
- help='Save fitted values of all variables as a snapshot and restore all variables to '
- 'their initial values. Should be used together with --save_ws.')
-@click.option('--rebuild/--no-rebuild', default=True, show_default=True,
- help='Save fitted workspace by rebuilding it. Should be used together with --save_ws.')
-@click.option('--outdir', default="pulls", show_default=True,
- help='Output directory for pulls output.')
-@click.option('--export_as_np_pulls/--skip_export_as_np_pulls', default=False, show_default=True,
- help='Export (constrained) NP results for pulls plot.')
-@click.option('-w', '--workspace', 'ws_name', default=None, show_default=True,
- help='Name of workspace. Auto-detect by default.')
-@click.option('-m', '--model_config', 'mc_name', default=None, show_default=True,
- help='Name of model config. Auto-detect by default.')
-@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
- help='Name of dataset.')
-@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
- help='Name of initial snapshot')
-@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
- help='Parameters to profile')
-@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
- help='Parameters to fix')
-@click.option('--pois', default="", show_default=True,
- help='Define the set of POIs (separated by commas) set for calculating Minos errors.')
-@click.option('--constrain/--no-constrain', 'constrain_nuis', default=True, show_default=True,
- help='Use constrained NLL (i.e. include systematics)')
-@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
- help='Minimizer type')
-@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
- help='Minimizer algorithm')
-@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
- help='Number of CPUs to use per parameter')
-@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
- help='Convergence criterium')
-@click.option('--retry', type=int, default=0, show_default=True,
- help='Maximum number of retries upon a failed fit')
-@click.option('--strategy', type=int, default=1, show_default=True,
- help='Default minimization strategy')
-@click.option('--hesse/--no-hesse', default=False, show_default=True,
- help='Evaluate errors using Hesse.')
-@click.option('--improve', type=int, default=0, show_default=True,
- help='Execute improve after each minimization')
-@click.option('--minimizer_offset', type=int, default=1, show_default=True,
- help='Enable minimizer offsetting')
-@click.option('--minos/--no-minos', default=False, show_default=True,
- help='Evaluate errors using Minos.')
-@click.option('--print_level', type=int, default=-1, show_default=True,
- help='Minimizer print level')
-@click.option('--fix-cache/--no-fix-cache', default=True, show_default=True,
- help='Fix StarMomentMorph cache')
-@click.option('--fix-multi/--no-fix-cache', default=True, show_default=True,
- help='Fix MultiPdf level 2')
-@click.option('--max_calls', type=int, default=-1, show_default=True,
- help='Maximum number of function calls')
-@click.option('--max_iters', type=int, default=-1, show_default=True,
- help='Maximum number of Minuit iterations')
-@click.option('--optimize', type=int, default=2, show_default=True,
- help='Optimize constant terms')
-@click.option('--offset/--no-offset', default=True, show_default=True,
- help='Offset likelihood')
-@click.option('--batch_mode/--no-batch', default=False, show_default=True,
- help='Batch mode when evaluating likelihood')
-@click.option('--int_bin_precision', type=float, default=-1., show_default=True,
- help='Integrate the PDF over the bins instead of using the probability '
- 'density at the bin centre')
-@click.option('-v', '--verbosity', default='INFO', show_default=True,
- type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
- help='Verbosity level.')
-def likelihood_fit(**kwargs):
- """
- Perform likelihood fit on a workspace
- """
- do_minos = kwargs.pop("minos")
- rebuild = kwargs.pop("rebuild")
- from quickstats.utils.string_utils import split_str
- pois = split_str(kwargs.pop("pois"), sep=',', remove_empty=True)
- _kwargs = {}
- for arg_name in ["outname", "save_log", "display", "save_result",
- "export_as_np_pulls", "outdir", "save_ws", "save_snapshot"]:
- _kwargs[arg_name] = kwargs.pop(arg_name)
- _init_kwargs = {}
- for arg_name in ["filename", "data_name", "verbosity"]:
- _init_kwargs[arg_name] = kwargs.pop(arg_name)
- _init_kwargs['config'] = kwargs
- _init_kwargs['poi_name'] = pois
- from quickstats.components import AnalysisBase
- if _kwargs['save_log']:
- from quickstats.concurrent.logging import standard_log
- log_path = os.path.splitext(_kwargs["outname"])[0] + ".log"
- with standard_log(log_path) as logger:
- analysis = AnalysisBase(**_init_kwargs)
- if _kwargs['export_as_np_pulls']:
- analysis.minimizer.configure(hesse=True)
- fit_result = analysis.nll_fit(mode=3, do_minos=do_minos)
- print(f"INFO: Saved fit log to `{log_path}`")
- else:
- analysis = AnalysisBase(**_init_kwargs)
- fit_result = analysis.nll_fit(mode=3, do_minos=do_minos)
- output = {}
- output['fit_result'] = fit_result
- df = {'pois':{}, 'nuisance_parameters':{}}
- analysis.load_snapshot("currentSnapshot")
- df['pois']['prefit'] = analysis.model.as_dataframe('poi')
- df['nuisance_parameters']['prefit'] = analysis.model.as_dataframe('nuisance_parameter')
- analysis.load_snapshot("nllFit")
- if do_minos:
- df['pois']['postfit'] = analysis.model.as_dataframe('poi', asym_error=True)
- else:
- df['pois']['postfit'] = analysis.model.as_dataframe('poi')
- df['nuisance_parameters']['postfit'] = analysis.model.as_dataframe('nuisance_parameter')
- if _kwargs['display']:
- import pandas as pd
- pd.set_option('display.max_rows', None)
- for key in ['pois', 'nuisance_parameters']:
- df[key]['combined'] = df[key]['prefit'].drop(["value", "error"], axis=1)
- df[key]['combined']['value_prefit'] = df[key]['prefit']['value']
- df[key]['combined']['value_postfit'] = df[key]['postfit']['value']
- df[key]['combined']['error_prefit'] = df[key]['prefit']['error']
- if (key == "pois") and do_minos:
- df[key]['combined']['errorlo_postfit'] = df[key]['postfit']['errorlo']
- df[key]['combined']['errorhi_postfit'] = df[key]['postfit']['errorhi']
- else:
- df[key]['combined']['error_postfit'] = df[key]['postfit']['error']
- output[key] = df[key]['combined'].to_dict("list")
- if _kwargs['display']:
- print("{}:".format(key.title()))
- print(df[key]['combined'])
- print()
- if _kwargs['save_result']:
- import json
- with open(_kwargs["outname"], "w") as f:
- json.dump(output, f, indent=2)
- print(f"INFO: Saved fit result to `{_kwargs['outname']}`")
- if _kwargs['export_as_np_pulls']:
- outdir = _kwargs['outdir']
- if not os.path.exists(outdir):
- os.makedirs(outdir)
- nuis_df = df[key]['combined'].drop(['min', 'max', 'is_constant', 'error_prefit'], axis=1)
- nuis_df = nuis_df.rename(columns={"value_prefit":"nuis_nom", "name":"nuisance",
- "value_postfit":"nuis_hat", "error_postfit":"nuis_hi"})
- nuis_df["nuis_lo"] = nuis_df["nuis_hi"]
- nuis_df["nuis_prefit"] = 1.0
- nuis_df = nuis_df.set_index(['nuisance'])
- constrained_np = [i.GetName() for i in analysis.model.get_constrained_nuisance_parameters()]
- nuis_df = nuis_df.loc[constrained_np].reset_index()
- nuis_data = nuis_df.to_dict('index')
- import json
- for i in nuis_data:
- data = nuis_data[i]
- np_name = data['nuisance']
- outpath = os.path.join(outdir, f"{np_name}.json")
- with open(outpath, "w") as outfile:
- json.dump({"nuis": data}, outfile, indent=2)
- if _kwargs["save_ws"] is not None:
- filename = _kwargs["save_ws"]
- if _kwargs["save_snapshot"] is not None:
- snapshot_name = _kwargs["save_snapshot"]
- from quickstats.components.basics import WSArgument
- analysis.save_snapshot(snapshot_name, WSArgument.MUTABLE)
- analysis.load_snapshot(analysis.kInitialSnapshotName)
- analysis.save(filename, rebuild=rebuild)
-
-@click.command(name='np_correlation')
-@click.option('-i', '--input_file', "filename", required=True,
- help='Path to the input workspace file.')
-@click.option('-o', '--basename', default='NP_correlation_matrix', show_default=True,
- help='Base name of the output.')
-@click.option('--select', default=None, show_default=True,
- help='Select specific NPs to be stored in the final output (for json and plot only). '
- 'Use comma to separate the selection (wild card is supported).')
-@click.option('--remove', default=None, show_default=True,
- help='Select specific NPs to be removed in the final output (for json and plot only). '
- 'Use comma to separate the selection (wild card is supported).')
-@click.option('--save_plot/--no_save_plot', default=True, show_default=True,
- help='Save NP correlation matrix as a plot in pdf format')
-@click.option('--save_json/--no_save_json', default=False, show_default=True,
- help='Save NP correlation matrix as a json file')
-@click.option('--save_root/--no_save_root', default=False, show_default=True,
- help='Save NP correlation matrix as a 2D histogram in a root file')
-@click.option('--plot_style', default="default", show_default=True,
- help='Plot style if save_plot is enabled. Choose between \"default\" and '
- f'\"viridis\". Alternatively, a path to a yaml config file can be used')
-@click.option('-w', '--workspace', 'ws_name', default=None, show_default=True,
- help='Name of workspace. Auto-detect by default.')
-@click.option('-m', '--model_config', 'mc_name', default=None, show_default=True,
- help='Name of model config. Auto-detect by default.')
-@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
- help='Name of dataset.')
-@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
- help='Name of initial snapshot')
-@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
- help='Parameters to profile')
-@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
- help='Parameters to fix')
-@click.option('--constrain/--no-constrain', 'constrain_nuis', default=True, show_default=True,
- help='Use constrained NLL (i.e. include systematics)')
-@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
- help='Minimizer type')
-@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
- help='Minimizer algorithm')
-@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
- help='Number of CPUs to use per parameter')
-@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
- help='Convergence criterium')
-@click.option('--strategy', type=int, default=1, show_default=True,
- help='Default minimization strategy')
-@click.option('--print_level', type=int, default=-1, show_default=True,
- help='Minimizer print level')
-@click.option('--fix-cache/--no-fix-cache', default=True, show_default=True,
- help='Fix StarMomentMorph cache')
-@click.option('--fix-multi/--no-fix-cache', default=True, show_default=True,
- help='Fix MultiPdf level 2')
-@click.option('--max_calls', type=int, default=-1, show_default=True,
- help='Maximum number of function calls')
-@click.option('--max_iters', type=int, default=-1, show_default=True,
- help='Maximum number of Minuit iterations')
-@click.option('--optimize', type=int, default=2, show_default=True,
- help='Optimize constant terms')
-@click.option('--improve', type=int, default=0, show_default=True,
- help='Execute improve after each minimization')
-@click.option('--minimizer_offset', type=int, default=1, show_default=True,
- help='Enable minimizer offsetting')
-@click.option('--offset/--no-offset', default=True, show_default=True,
- help='Offset likelihood')
-@click.option('--batch_mode/--no-batch', default=False, show_default=True,
- help='Batch mode when evaluating likelihood')
-@click.option('--int_bin_precision', type=float, default=-1., show_default=True,
- help='Integrate the PDF over the bins instead of using the probability '
- 'density at the bin centre')
-@click.option('-v', '--verbosity', default='INFO', show_default=True,
- type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
- help='Verbosity level.')
-def np_correlation(**kwargs):
- """
- Evaluate post-fit NP correlation matrix
- """
- _kwargs = {}
- for arg_name in ["basename", "save_plot", "save_json", "save_root",
- "plot_style", "select", "remove"]:
- _kwargs[arg_name] = kwargs.pop(arg_name)
- _init_kwargs = {}
- for arg_name in ["filename", "data_name", "verbosity"]:
- _init_kwargs[arg_name] = kwargs.pop(arg_name)
- _init_kwargs['config'] = kwargs
- _init_kwargs['poi_name'] = []
- from quickstats.components import AnalysisBase
- analysis = AnalysisBase(**_init_kwargs)
- analysis.minimizer.configure(hesse=True)
- analysis.nll_fit(mode=3)
- fit_result = analysis.roofit_result
- basename = os.path.splitext(_kwargs['basename'])[0]
- from quickstats.utils.roofit_utils import get_correlation_matrix
- if _kwargs['save_root']:
- correlation_hist = get_correlation_matrix(fit_result, lib="root")
- outname = basename + ".root"
- correlation_hist.SaveAs(outname)
- print(f"INFO: Saved correlation histogram to `{outname}`")
- correlation_hist.Delete()
- from quickstats.utils.common_utils import filter_by_wildcards
- if _kwargs['save_json'] or _kwargs['save_plot']:
- df = get_correlation_matrix(fit_result, lib='pandas')
- labels = list(df.columns)
- selected = filter_by_wildcards(labels, _kwargs['select'])
- selected = filter_by_wildcards(selected, _kwargs['remove'], exclusion=True)
- to_drop = list(set(labels) - set(selected))
- df = df.drop(to_drop, axis=0).drop(to_drop, axis=1).transpose()
- if _kwargs['save_json']:
- data = df.to_dict()
- outname = basename + ".json"
- with open(outname, "w") as out:
- json.dump(data, out, indent=2)
- print(f"INFO: Saved correlation data to `{outname}`")
- if _kwargs['save_plot']:
- import matplotlib.pyplot as plt
- from quickstats.plots import CorrelationPlot
- plotter = CorrelationPlot(df)
- ax = plotter.draw_style(_kwargs['plot_style'])
- outname = basename + ".pdf"
- plt.savefig(outname, bbox_inches="tight")
- print(f"INFO: Saved correlation plot to `{outname}`")
\ No newline at end of file
diff --git a/quickstats/clis/likelihood_scan.py b/quickstats/clis/likelihood_scan.py
deleted file mode 100644
index cf9909c420c76b3cca1f97cfd07844851d68b266..0000000000000000000000000000000000000000
--- a/quickstats/clis/likelihood_scan.py
+++ /dev/null
@@ -1,130 +0,0 @@
-import os
-import click
-
-@click.command(name='likelihood_scan')
-@click.option('-i', '--input_file', required=True,
- help='Path to the input workspace file.')
-@click.option('-p', '--param_expr', default=None,
- help='\b\n Parameter expression, e.g.'
- '\b\n 1D scan: "poi_name=<poi_min>_<poi_max>_<step>"'
- '\b\n 2D scan: "poi_1_name=<poi_1_min>_<poi_1_max>_<step_1>,'
- '\b\n poi_2_name=<poi_2_min>_<poi_2_max>_<step_2>"')
-@click.option('--filter', 'filter_expr', default=None, show_default=True,
- help='\b Filter parameter points by expression.\n'
- '\b Example: "mass=2*,350,400,450;klambda=1.*,2.*,-1.*,-2.*"\n'
- '\b Refer to documentation for more information\n')
-@click.option('--exclude', 'exclude_expr', default=None, show_default=True,
- help='\b Exclude parameter points by expression.\n'
- '\b Example: "mass=2*,350,400,450;klambda=1.*,2.*,-1.*,-2.*"\n'
- '\b Refer to documentation for more information\n')
-@click.option('--min', 'poi_min', type=float, default=None,
- help='(deprecated) Minimum POI value to scan.')
-@click.option('--max', 'poi_max', type=float, default=None,
- help='(deprecated) Maximum POI value to scan.')
-@click.option('--step', 'poi_step', type=float, default=None,
- help='(deprecated) Scan interval.')
-@click.option('--poi', 'poi_name', default=None, show_default=True,
- help='(deprecated) POI to scan. If not specified, the first POI from the workspace is used.')
-@click.option('--cache/--no-cache', default=True, show_default=True,
- help='Cache existing result.')
-@click.option('-o', '--outname', default='{poi_names}.json', show_default=True,
- help='Name of output')
-@click.option('--outdir', default='likelihood_scan', show_default=True,
- help='Output directory.')
-@click.option('--cachedir', default='cache', show_default=True,
- help='Cache directory relative to the output directory.')
-@click.option('--save_log/--skip_log', default=True, show_default=True,
- help='Save log file.')
-@click.option('-w', '--workspace', 'ws_name', default=None, show_default=True,
- help='Name of workspace. Auto-detect by default.')
-@click.option('-m', '--model_config', 'mc_name', default=None, show_default=True,
- help='Name of model config. Auto-detect by default.')
-@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
- help='Name of dataset.')
-@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
- help='Name of initial snapshot.')
-@click.option('--uncond_snapshot', default=None, show_default=True,
- help='Name of snapshot with unconditional fit result.')
-@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
- help='Parameters to profile.')
-@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
- help='Parameters to fix.')
-@click.option('--constrain/--no-constrain', 'constrain_nuis', default=True, show_default=True,
- help='Use constrained NLL (i.e. include systematics).')
-@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
- help='Minimizer type.')
-@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
- help='Minimizer algorithm.')
-@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
- help='Number of CPUs to use per parameter.')
-@click.option('--binned/--unbinned', 'binned_likelihood', default=True, show_default=True,
- help='Binned likelihood.')
-@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
- help='Convergence criterium.')
-@click.option('--retry', type=int, default=0, show_default=True,
- help='Maximum number of retries upon a failed fit.')
-@click.option('--strategy', type=int, default=1, show_default=True,
- help='Default minimization strategy.')
-@click.option('--print_level', type=int, default=-1, show_default=True,
- help='Minimizer print level.')
-@click.option('--fix-cache/--no-fix-cache', default=True, show_default=True,
- help='Fix StarMomentMorph cache.')
-@click.option('--fix-multi/--no-fix-cache', default=True, show_default=True,
- help='Fix MultiPdf level 2.')
-@click.option('--max_calls', type=int, default=-1, show_default=True,
- help='Maximum number of function calls.')
-@click.option('--max_iters', type=int, default=-1, show_default=True,
- help='Maximum number of Minuit iterations.')
-@click.option('--optimize', type=int, default=2, show_default=True,
- help='Optimize constant terms.')
-@click.option('--improve', type=int, default=0, show_default=True,
- help='Execute improve after each minimization')
-@click.option('--minimizer_offset', type=int, default=1, show_default=True,
- help='Enable minimizer offsetting')
-@click.option('--offset/--no-offset', default=True, show_default=True,
- help='Offset likelihood.')
-@click.option('--allow-nan/--not-allow-nan', default=True, show_default=True,
- help='Allow cached nll to be nan.')
-@click.option('-v', '--verbosity', default='INFO', show_default=True,
- type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
- help='Verbosity level.')
-@click.option('--parallel', type=int, default=-1, show_default=True,
- help='\b\n Parallelize job across the N workers.'
- '\b\n Case 0: Jobs are run sequentially (for debugging).'
- '\b\n Case -1: Jobs are run across N_CPU workers.')
-def likelihood_scan(**kwargs):
- """
- Evaluate a set of parmeterised likelihood values
- """
- _kwargs = {}
- for arg_name in ["input_file", "param_expr", "data_name", "outdir", "filter_expr", "uncond_snapshot",
- "exclude_expr", "outname", "cache", "cachedir", "save_log", "parallel", "verbosity",
- "allow_nan"]:
- _kwargs[arg_name] = kwargs.pop(arg_name)
- _kwargs['config'] = kwargs
-
- # for backward compatibility
- _deprecated_kwargs = {}
- for arg_name in ["poi_min", "poi_max", "poi_step", "poi_name"]:
- _deprecated_kwargs[arg_name] = kwargs.pop(arg_name)
- if all(_deprecated_kwargs[arg_name] is not None for arg_name in ["poi_min", "poi_max", "poi_step"]):
- if _kwargs['param_expr'] is not None:
- raise RuntimeError("either `param_expr` or (`poi_min`, `poi_max`, `poi_step`) should be "
- "given for 1D likelihood scan")
- print("WARNING: Likelihood scan using `poi_min`, `poi_max` and `poi_step` is "
- "deprecated. Use --param_expr \"<poi_name>=<poi_min>_<poi_max>_<poi_step>\" instead.")
- from quickstats.components import AnalysisBase
- analysis = AnalysisBase(_kwargs['input_file'], poi_name=_deprecated_kwargs['poi_name'],
- data_name=_kwargs['data_name'], verbosity="WARNING")
- poi_name = analysis.poi.GetName()
- poi_min = _deprecated_kwargs['poi_min']
- poi_max = _deprecated_kwargs['poi_max']
- poi_step = _deprecated_kwargs['poi_step']
- _kwargs['param_expr'] = f"{poi_name}={poi_min}_{poi_max}_{poi_step}"
- elif (not all(_deprecated_kwargs[arg_name] is None for arg_name in ["poi_min", "poi_max", "poi_step"])) or \
- (_kwargs['param_expr'] is None):
- raise RuntimeError("either `param_expr` or (`poi_min`, `poi_max`, `poi_step`) should be "
- "given for 1D likelihood scan")
- from quickstats.concurrent import ParameterisedLikelihood
- runner = ParameterisedLikelihood(**_kwargs)
- runner.run()
diff --git a/quickstats/clis/likelihood_tools.py b/quickstats/clis/likelihood_tools.py
new file mode 100644
index 0000000000000000000000000000000000000000..39eed9ce5c9079bfb9b3471949cf8bc17abf568b
--- /dev/null
+++ b/quickstats/clis/likelihood_tools.py
@@ -0,0 +1,394 @@
+import os
+import json
+import click
+
+from .core import cli
+
+__all__ = ['likelihood_fit', 'likelihood_scan']
+
+@cli.command(name='likelihood_fit')
+@click.option('-i', '--input_file', "filename", required=True,
+ help='Path to the input workspace file.')
+@click.option('-o', '--outname', default='fit_result.json', show_default=True,
+ help='Name of output file.')
+@click.option('--display/--no-display', default=True, show_default=True,
+ help='Display fit result.')
+@click.option('--save/--no-save', "save_result", default=False, show_default=True,
+ help='Save fit result.')
+@click.option('--save_log/--skip_log', default=False, show_default=True,
+ help='Save log file.')
+@click.option('--save_ws', default=None, show_default=True,
+ help='Save fitted workspace to a given path.')
+@click.option('--save_snapshot', default=None, show_default=True,
+ help='Save fitted values of all variables as a snapshot and restore all variables to '
+ 'their initial values. Should be used together with --save_ws.')
+@click.option('--rebuild/--no-rebuild', default=True, show_default=True,
+ help='Save fitted workspace by rebuilding it. Should be used together with --save_ws.')
+@click.option('--export_as_np_pulls/--skip_export_as_np_pulls', default=False, show_default=True,
+ help='Export (constrained) NP results for pulls plot.')
+@click.option('--outdir', default="pulls", show_default=True,
+ help='Output directory for pulls output.')
+@click.option('-w', '--workspace', 'ws_name', default=None, show_default=True,
+ help='Name of workspace. Auto-detect by default.')
+@click.option('-m', '--model_config', 'mc_name', default=None, show_default=True,
+ help='Name of model config. Auto-detect by default.')
+@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
+ help='Name of dataset.')
+@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
+ help='Name of initial snapshot.')
+@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
+ help='Parameters to profile.')
+@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
+ help='Parameters to fix.')
+@click.option('--pois', default="", show_default=True,
+ help='Define the set of POIs (separated by commas) set for calculating Minos errors.')
+@click.option('--constrain/--no-constrain', 'constrain_nuis', default=True, show_default=True,
+ help='Use constrained NLL (i.e. include systematics).')
+@click.option('--minos/--no-minos', default=False, show_default=True,
+ help='Evaluate errors using Minos.')
+@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
+ help='Minimizer type.')
+@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
+ help='Minimizer algorithm.')
+@click.option('--strategy', type=int, default=1, show_default=True,
+ help='Default minimization strategy.')
+@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
+ help='Minimization convergence criterium.')
+@click.option('--retry', type=int, default=1, show_default=True,
+ help='Maximum number of retries upon a failed fit.')
+@click.option('--optimize', type=int, default=2, show_default=True,
+ help='Optimize constant terms.')
+@click.option('--minimizer_offset', type=int, default=1, show_default=True,
+ help='Enable minimizer offsetting.')
+@click.option('--offset/--no-offset', default=True, show_default=True,
+ help='Offset likelihood.')
+@click.option('--binned/--unbinned', 'binned_likelihood', default=True, show_default=True,
+ help='Activate binned likelihood for RooRealSumPdf.')
+@click.option('--print_level', type=int, default=-1, show_default=True,
+ help='Minimizer print level.')
+@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
+ help='Number of CPUs to use during minimization.')
+@click.option('--batch_mode/--no-batch', default=False, show_default=True,
+ help='Batch mode when evaluating likelihood.')
+@click.option('--int_bin_precision', type=float, default=-1., show_default=True,
+ help='Integrate the PDF over the bins instead of using the probability '
+ 'density at the bin center.')
+@click.option('--extra_minimizer_options', default=None, show_default=True,
+ help='Additional minimizer options to include. Format should be <config>=<value> '
+ 'separated by commas. Example: "discrete_min_tol=0.001,do_discrete_iteration=1"')
+@click.option('--cms_runtimedef', 'runtimedef_expr', default=None, show_default=True,
+ help='CMS specific runtime definitions. Format should be <config>=<value> '
+ 'separated by commas. Example: "REMOVE_CONSTANT_ZERO_POINT=1,ADDNLL_GAUSSNLL=0"')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
+def likelihood_fit(**kwargs):
+ """
+ Perform likelihood fit on a workspace
+ """
+ do_minos = kwargs.pop("minos")
+ rebuild = kwargs.pop("rebuild")
+ from quickstats.utils.string_utils import split_str
+ pois = split_str(kwargs.pop("pois"), sep=',', remove_empty=True)
+ _kwargs = {}
+ for arg_name in ["outname", "save_log", "display", "save_result",
+ "export_as_np_pulls", "outdir", "save_ws", "save_snapshot"]:
+ _kwargs[arg_name] = kwargs.pop(arg_name)
+ _init_kwargs = {}
+ for arg_name in ["filename", "data_name", "verbosity"]:
+ _init_kwargs[arg_name] = kwargs.pop(arg_name)
+ _init_kwargs['config'] = kwargs
+ _init_kwargs['poi_name'] = pois
+ from quickstats.components import AnalysisBase
+ if _kwargs['save_log']:
+ from quickstats.concurrent.logging import standard_log
+ log_path = os.path.splitext(_kwargs["outname"])[0] + ".log"
+ with standard_log(log_path) as logger:
+ analysis = AnalysisBase(**_init_kwargs)
+ if _kwargs['export_as_np_pulls']:
+ analysis.minimizer.configure(hesse=True)
+ fit_result = analysis.nll_fit(mode=3, do_minos=do_minos)
+ print(f"INFO: Saved fit log to `{log_path}`")
+ else:
+ analysis = AnalysisBase(**_init_kwargs)
+ fit_result = analysis.nll_fit(mode=3, do_minos=do_minos)
+ output = {}
+ output['fit_result'] = fit_result
+ df = {'pois':{}, 'nuisance_parameters':{}}
+ analysis.load_snapshot("currentSnapshot")
+ df['pois']['prefit'] = analysis.model.as_dataframe('poi')
+ df['nuisance_parameters']['prefit'] = analysis.model.as_dataframe('nuisance_parameter')
+ analysis.load_snapshot("nllFit")
+ if do_minos:
+ df['pois']['postfit'] = analysis.model.as_dataframe('poi', asym_error=True)
+ else:
+ df['pois']['postfit'] = analysis.model.as_dataframe('poi')
+ df['nuisance_parameters']['postfit'] = analysis.model.as_dataframe('nuisance_parameter')
+ if _kwargs['display']:
+ import pandas as pd
+ pd.set_option('display.max_rows', None)
+ for key in ['pois', 'nuisance_parameters']:
+ df[key]['combined'] = df[key]['prefit'].drop(["value", "error"], axis=1)
+ df[key]['combined']['value_prefit'] = df[key]['prefit']['value']
+ df[key]['combined']['value_postfit'] = df[key]['postfit']['value']
+ df[key]['combined']['error_prefit'] = df[key]['prefit']['error']
+ if (key == "pois") and do_minos:
+ df[key]['combined']['errorlo_postfit'] = df[key]['postfit']['errorlo']
+ df[key]['combined']['errorhi_postfit'] = df[key]['postfit']['errorhi']
+ else:
+ df[key]['combined']['error_postfit'] = df[key]['postfit']['error']
+ output[key] = df[key]['combined'].to_dict("list")
+ if _kwargs['display']:
+ print("{}:".format(key.title()))
+ print(df[key]['combined'])
+ print()
+ if _kwargs['save_result']:
+ import json
+ with open(_kwargs["outname"], "w") as f:
+ json.dump(output, f, indent=2)
+ print(f"INFO: Saved fit result to `{_kwargs['outname']}`")
+ if _kwargs['export_as_np_pulls']:
+ outdir = _kwargs['outdir']
+ if not os.path.exists(outdir):
+ os.makedirs(outdir)
+ nuis_df = df[key]['combined'].drop(['min', 'max', 'is_constant', 'error_prefit'], axis=1)
+ nuis_df = nuis_df.rename(columns={"value_prefit":"nuis_nom", "name":"nuisance",
+ "value_postfit":"nuis_hat", "error_postfit":"nuis_hi"})
+ nuis_df["nuis_lo"] = nuis_df["nuis_hi"]
+ nuis_df["nuis_prefit"] = 1.0
+ nuis_df = nuis_df.set_index(['nuisance'])
+ constrained_np = [i.GetName() for i in analysis.model.get_constrained_nuisance_parameters()]
+ nuis_df = nuis_df.loc[constrained_np].reset_index()
+ nuis_data = nuis_df.to_dict('index')
+ import json
+ for i in nuis_data:
+ data = nuis_data[i]
+ np_name = data['nuisance']
+ outpath = os.path.join(outdir, f"{np_name}.json")
+ with open(outpath, "w") as outfile:
+ json.dump({"nuis": data}, outfile, indent=2)
+ if _kwargs["save_ws"] is not None:
+ filename = _kwargs["save_ws"]
+ if _kwargs["save_snapshot"] is not None:
+ snapshot_name = _kwargs["save_snapshot"]
+ from quickstats.components.basics import WSArgument
+ analysis.save_snapshot(snapshot_name, WSArgument.MUTABLE)
+ analysis.load_snapshot(analysis.kInitialSnapshotName)
+ analysis.save(filename, rebuild=rebuild)
+
+@cli.command(name='likelihood_scan')
+@click.option('-i', '--input_path', required=True,
+ help='Input directory/path containing the workspace file(s) to process.')
+@click.option('--file_expr', default=None, show_default=True,
+ help='\b\n File name expression describing the external parameterisation.'
+ '\b\n Example: "<mass[F]>_kl_<klambda[P]>"'
+ '\b\n Regular expression is supported'
+ '\b\n Refer to documentation for more information')
+@click.option('-p', '--param_expr', default=None,
+ help='\b\n Parameter expression, e.g.'
+ '\b\n 1D scan: "poi_name=<poi_min>_<poi_max>_<step>"'
+ '\b\n 2D scan: "poi_1_name=<poi_1_min>_<poi_1_max>_<step_1>,'
+ '\b\n poi_2_name=<poi_2_min>_<poi_2_max>_<step_2>"')
+@click.option('--filter', 'filter_expr', default=None, show_default=True,
+ help='\b\n Filter parameter points by expression.'
+ '\b\n Example: "mass=2*,350,400,450;klambda=1.*,2.*,-1.*,-2.*"'
+ '\b\n Refer to documentation for more information')
+@click.option('--exclude', 'exclude_expr', default=None, show_default=True,
+ help='\b\n Exclude parameter points by expression.'
+ '\b\n Example: "mass=2*,350,400,450;klambda=1.*,2.*,-1.*,-2.*"'
+ '\b\n Refer to documentation for more information')
+@click.option('--cache/--no-cache', default=True, show_default=True,
+ help='Cache existing result.')
+@click.option('-o', '--outname', default='{poi_names}.json', show_default=True,
+ help='Name of output file.')
+@click.option('--outdir', default='likelihood_scan', show_default=True,
+ help='Output directory.')
+@click.option('--cachedir', default='cache', show_default=True,
+ help='Cache directory relative to the output directory.')
+@click.option('--save_log/--skip_log', default=True, show_default=True,
+ help='Save log file.')
+@click.option('-w', '--workspace', 'ws_name', default=None, show_default=True,
+ help='Name of workspace. Auto-detect by default.')
+@click.option('-m', '--model_config', 'mc_name', default=None, show_default=True,
+ help='Name of model config. Auto-detect by default.')
+@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
+ help='Name of dataset.')
+@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
+ help='Name of initial snapshot.')
+@click.option('--uncond_snapshot', default=None, show_default=True,
+ help='Name of snapshot with unconditional fit result.')
+@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
+ help='Parameters to profile.')
+@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
+ help='Parameters to fix.')
+@click.option('--constrain/--no-constrain', 'constrain_nuis', default=True, show_default=True,
+ help='Use constrained NLL (i.e. include systematics).')
+@click.option('--allow-nan/--not-allow-nan', default=True, show_default=True,
+ help='Allow cached nll to be nan.')
+@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
+ help='Minimizer type.')
+@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
+ help='Minimizer algorithm.')
+@click.option('--strategy', type=int, default=1, show_default=True,
+ help='Default minimization strategy.')
+@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
+ help='Minimization convergence criterium.')
+@click.option('--retry', type=int, default=1, show_default=True,
+ help='Maximum number of retries upon a failed fit.')
+@click.option('--optimize', type=int, default=2, show_default=True,
+ help='Optimize constant terms.')
+@click.option('--minimizer_offset', type=int, default=1, show_default=True,
+ help='Enable minimizer offsetting.')
+@click.option('--offset/--no-offset', default=True, show_default=True,
+ help='Offset likelihood.')
+@click.option('--binned/--unbinned', 'binned_likelihood', default=True, show_default=True,
+ help='Activate binned likelihood for RooRealSumPdf.')
+@click.option('--print_level', type=int, default=-1, show_default=True,
+ help='Minimizer print level.')
+@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
+ help='Number of CPUs to use during minimization.')
+@click.option('--batch_mode/--no-batch', default=False, show_default=True,
+ help='Batch mode when evaluating likelihood.')
+@click.option('--int_bin_precision', type=float, default=-1., show_default=True,
+ help='Integrate the PDF over the bins instead of using the probability '
+ 'density at the bin center.')
+@click.option('--extra_minimizer_options', default=None, show_default=True,
+ help='Additional minimizer options to include. Format should be <config>=<value> '
+ 'separated by commas. Example: "discrete_min_tol=0.001,do_discrete_iteration=1"')
+@click.option('--cms_runtimedef', 'runtimedef_expr', default=None, show_default=True,
+ help='CMS specific runtime definitions. Format should be <config>=<value> '
+ 'separated by commas. Example: "REMOVE_CONSTANT_ZERO_POINT=1,ADDNLL_GAUSSNLL=0"')
+@click.option('--parallel', type=int, default=-1, show_default=True,
+ help='\b\n Parallelize job across the N workers.'
+ '\b\n Case 0: Jobs are run sequentially (for debugging).'
+ '\b\n Case -1: Jobs are run across N_CPU workers.')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
+def likelihood_scan(**kwargs):
+ """
+ Evaluate a set of parmeterised likelihood values
+ """
+ _kwargs = {}
+ for arg_name in ["input_path", "file_expr", "param_expr", "data_name", "outdir", "filter_expr",
+ "uncond_snapshot", "exclude_expr", "outname", "cache", "cachedir", "save_log",
+ "parallel", "verbosity", "allow_nan"]:
+ _kwargs[arg_name] = kwargs.pop(arg_name)
+ _kwargs['config'] = kwargs
+ from quickstats.concurrent import ParameterisedLikelihood
+ runner = ParameterisedLikelihood(**_kwargs)
+ runner.run()
+
+@cli.command(name='significance_scan')
+@click.option('-i', '--input_path', required=True,
+ help='Path to the input workspace file or directory containing the parameterised '
+ 'input workspace files.')
+@click.option('-p', '--poi', 'poi_name', default=None,
+ help='Name of the parameter of interest (POI). If None, the first POI is used.')
+@click.option('--mu_exp', type=float, default=0., show_default=True,
+ help='Expected value of the POI under the null hypothesis.')
+@click.option('--asimov_type', type=int, default=None,
+ help='\b\n Evaluate significance on an Asimov dataset of this type. '
+ 'If not specified, the observed data is used. '
+ '\b\n Choices of asimov types are'
+ '\b\n 0: fit with POI fixed to 0'
+ '\b\n 1: fit with POI fixed to 1'
+ '\b\n 2: fit with POI free and set POI to 1 after fit'
+ '\b\n 3: fit with POI and constrained NP fixed to 0'
+ '\b\n 4: fit with POI fixed to 1 and constrained NP fixed to 0'
+ '\b\n 5: fit with POI free and constrained NP fixed to 0 and set POI to 1 after fit'
+ '\b\n -1: nominal NP with POI set to 0'
+ '\b\n -2: nominal NP with POI set to 1')
+@click.option('--file_expr', default=r"[\w-]+", show_default=True,
+ help='\b\n File name expression describing the external parameterisation.'
+ '\b\n Example: "<mass[F]>_kl_<klambda[P]>"'
+ '\b\n Regular expression is supported'
+ '\b\n Refer to documentation for more information')
+@click.option('--param_expr', default=None, show_default=True,
+ help='\b\n Parameter name expression describing the internal parameterisation.'
+ '\b\n Example: "klambda=-10_10_0.2,k2v=(1,2,3)"'
+ '\b\n Refer to documentation for more information')
+@click.option('--filter', 'filter_expr', default=None, show_default=True,
+ help='\b\n Filter parameter points by expression.'
+ '\b\n Example: "mass=(2*,350,400,450);klambda=(1.*,2.*,-1.*,-2.*)"'
+ '\b\n Refer to documentation for more information')
+@click.option('--exclude', 'exclude_expr', default=None, show_default=True,
+ help='\b\n Exclude parameter points by expression.'
+ '\b\n Example: "mass=(2*,350,400,450);klambda=(1.*,2.*,-1.*,-2.*)"'
+ '\b\n Refer to documentation for more information')
+@click.option('--outdir', default='significance', show_default=True,
+ help='Output directory where cached limit files and the merged limit file are saved.')
+@click.option('--cachedir', default='cache', show_default=True,
+ help='Cache directory relative to the output directory.')
+@click.option('--cache/--no-cache', default=True, show_default=True,
+ help='Cache existing result.')
+@click.option('-o', '--outname', default='{param_names}.json', show_default=True,
+ help='Name of the output significance file (all parameter points merged).')
+@click.option('--save_log/--skip_log', default=True, show_default=True,
+ help='Save log file.')
+@click.option('-w', '--workspace', 'ws_name', default=None, show_default=True,
+ help='Name of workspace. Auto-detect by default.')
+@click.option('-m', '--model_config', 'mc_name', default=None, show_default=True,
+ help='Name of model config. Auto-detect by default.')
+@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
+ help='Name of dataset.')
+@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
+ help='Name of initial snapshot.')
+@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
+ help='Parameters to profile.')
+@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
+ help='Parameters to fix.')
+@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
+ help='Minimizer type.')
+@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
+ help='Minimizer algorithm.')
+@click.option('--strategy', type=int, default=1, show_default=True,
+ help='Default minimization strategy.')
+@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
+ help='Minimization convergence criterium.')
+@click.option('--retry', type=int, default=1, show_default=True,
+ help='Maximum number of retries upon a failed fit.')
+@click.option('--optimize', type=int, default=2, show_default=True,
+ help='Optimize constant terms.')
+@click.option('--minimizer_offset', type=int, default=1, show_default=True,
+ help='Enable minimizer offsetting.')
+@click.option('--offset/--no-offset', default=True, show_default=True,
+ help='Offset likelihood.')
+@click.option('--binned/--unbinned', 'binned_likelihood', default=True, show_default=True,
+ help='Activate binned likelihood for RooRealSumPdf.')
+@click.option('--print_level', type=int, default=-1, show_default=True,
+ help='Minimizer print level.')
+@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
+ help='Number of CPUs to use during minimization.')
+@click.option('--batch_mode/--no-batch', default=False, show_default=True,
+ help='Batch mode when evaluating likelihood.')
+@click.option('--int_bin_precision', type=float, default=-1., show_default=True,
+ help='Integrate the PDF over the bins instead of using the probability '
+ 'density at the bin center.')
+@click.option('--extra_minimizer_options', default=None, show_default=True,
+ help='Additional minimizer options to include. Format should be <config>=<value> '
+ 'separated by commas. Example: "discrete_min_tol=0.001,do_discrete_iteration=1"')
+@click.option('--cms_runtimedef', 'runtimedef_expr', default=None, show_default=True,
+ help='CMS specific runtime definitions. Format should be <config>=<value> '
+ 'separated by commas. Example: "REMOVE_CONSTANT_ZERO_POINT=1,ADDNLL_GAUSSNLL=0"')
+@click.option('--parallel', type=int, default=-1, show_default=True,
+ help='\b\n Parallelize job across the N workers.'
+ '\b\n Case 0: Jobs are run sequentially (for debugging).'
+ '\b\n Case -1: Jobs are run across N_CPU workers.')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
+def significance_scan(**kwargs):
+ """
+ Evaluate a set of parmeterised significance values
+ """
+ _kwargs = {}
+ for arg_name in ["input_path", "poi_name", "data_name", "file_expr", "param_expr",
+ "filter_expr", "exclude_expr", "mu_exp", "asimov_type",
+ "snapshot_name", "outdir", "cachedir", "outname", "cache",
+ "save_log", "parallel", "verbosity"]:
+ _kwargs[arg_name] = kwargs.pop(arg_name)
+ _kwargs['config'] = kwargs
+ from quickstats.concurrent import ParameterisedSignificance
+ runner = ParameterisedSignificance(**_kwargs)
+ runner.run()
\ No newline at end of file
diff --git a/quickstats/clis/limit_setting.py b/quickstats/clis/limit_setting.py
index c5058df09765adca8c3115dc7314246f78ed7eca..7468606f08b3f35e0540fb43dd9c8cba1ddcb1bb 100644
--- a/quickstats/clis/limit_setting.py
+++ b/quickstats/clis/limit_setting.py
@@ -1,7 +1,103 @@
import os
import click
+
+from .core import cli
+
+__all__ = ['cls_limit', 'limit_scan']
+
+@cli.command(name='cls_limit')
+@click.option('-i', '--input_file', 'filename', required=True,
+ help='Path to the input workspace file.')
+@click.option('-p', '--poi', 'poi_name', default=None, show_default=True,
+ help='Name of the parameter of interest (POI). If not specified, the first POI from the workspace is used.')
+@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
+ help='Name of dataset.')
+@click.option('--asimov_data_name', 'asimov_data_name', default=None,
+ help='If given, use custom background asimov dataset instead of generating on the fly.')
+@click.option('-o', '--outname', default='limits.json', show_default=True,
+ help='Name of output limit file.')
+@click.option('--mu_exp', type=float, default=0, show_default=True,
+ help='(DO NOT USE) Expected signal strengh value to be used for Asimov generation.')
+@click.option('--blind/--unblind', 'do_blind', default=True, show_default=True,
+ help='Blind/unblind analysis.')
+@click.option('--CL', 'CL', type=float, default=0.95, show_default=True,
+ help='CL value to use.')
+@click.option('--precision', default=0.005, show_default=True,
+ help='precision in mu that defines iterative cutoff.')
+@click.option('--adjust_fit_range/--keep_fit_range', default=True, show_default=True,
+ help='whether to adjust the fit range to median limit +- 5 sigma for observed fit.')
+@click.option('--do_tilde/--no_tilde', default=True, show_default=True,
+ help='bound mu at zero if true and do the \tilde{q}_{mu} asymptotics.')
+@click.option('--predictive_fit/--no_predictive_fit', default=False, show_default=True,
+ help='extrapolate best fit nuisance parameters based on previous fit results.')
+@click.option('--do_better_bands/--skip_better_bands', default=True, show_default=True,
+ help='evaluate asymptotic CLs limit for various sigma bands.')
+@click.option('--better_negative_bands/--skip_better_negative_bands', default=False, show_default=True,
+ help='evaluate asymptotic CLs limit for negative sigma bands.')
+@click.option('--binned/--unbinned', 'binned_likelihood', default=True, show_default=True,
+ help='Activate binned likelihood for RooRealSumPdf.')
+@click.option('--save_summary/--skip_summary', default=True, show_default=True,
+ help='Save summary information.')
+@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
+ help='Parameters to fix.')
+@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
+ help='Parameters to profile.')
+@click.option('-w', '--workspace', 'ws_name', default=None, show_default=True,
+ help='Name of workspace. Auto-detect by default.')
+@click.option('-m', '--model_config', 'mc_name', default=None, show_default=True,
+ help='Name of model config. Auto-detect by default.')
+@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
+ help='Name of initial snapshot.')
+@click.option('--constrain/--no-constrain', 'constrain_nuis', default=True, show_default=True,
+ help='Use constrained NLL.')
+@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
+ help='Minimizer type.')
+@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
+ help='Minimizer algorithm.')
+@click.option('--strategy', type=int, default=1, show_default=True,
+ help='Default minimization strategy.')
+@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
+ help='Minimization convergence criterium.')
+@click.option('--retry', type=int, default=2, show_default=True,
+ help='Maximum number of retries upon a failed fit.')
+@click.option('--optimize', type=int, default=2, show_default=True,
+ help='Optimize constant terms.')
+@click.option('--minimizer_offset', type=int, default=1, show_default=True,
+ help='Enable minimizer offsetting.')
+@click.option('--offset/--no-offset', default=True, show_default=True,
+ help='Offset likelihood.')
+@click.option('--binned/--unbinned', 'binned_likelihood', default=True, show_default=True,
+ help='Activate binned likelihood for RooRealSumPdf.')
+@click.option('--print_level', type=int, default=-1, show_default=True,
+ help='Minimizer print level.')
+@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
+ help='Number of CPUs to use during minimization.')
+@click.option('--batch_mode/--no-batch', default=False, show_default=True,
+ help='Batch mode when evaluating likelihood.')
+@click.option('--int_bin_precision', type=float, default=-1., show_default=True,
+ help='Integrate the PDF over the bins instead of using the probability '
+ 'density at the bin center.')
+@click.option('--extra_minimizer_options', default=None, show_default=True,
+ help='Additional minimizer options to include. Format should be <config>=<value> '
+ 'separated by commas. Example: "discrete_min_tol=0.001,do_discrete_iteration=1"')
+@click.option('--cms_runtimedef', 'runtimedef_expr', default=None, show_default=True,
+ help='CMS specific runtime definitions. Format should be <config>=<value> '
+ 'separated by commas. Example: "REMOVE_CONSTANT_ZERO_POINT=1,ADDNLL_GAUSSNLL=0"')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
+def cls_limit(**kwargs):
+ """
+ Tool for evaluating Asymptotic CLs limit
+ """
+ from quickstats.components import AsymptoticCLs
+ outname = kwargs.pop('outname')
+ save_summary = kwargs.pop('save_summary')
+ asymptotic_cls = AsymptoticCLs(**kwargs)
+ asymptotic_cls.evaluate_limits()
+ asymptotic_cls.save(outname, summary=save_summary)
-@click.command(name='limit_scan')
+@cli.command(name='limit_scan')
@click.option('-i', '--input_path', 'input_path', required=True,
help='Input directory/path containing the workspace file(s) to process.')
@click.option('--file_expr', default=r"[\w-]+", show_default=True,
@@ -11,113 +107,110 @@ import click
'\b\n Refer to documentation for more information')
@click.option('--param_expr', default=None, show_default=True,
help='\b\n Parameter name expression describing the internal parameterisation.'
- '\b\n\n Example: "klambda=-10_10_0.2,k2v=1"'
+ '\b\n Example: "klambda=-10_10_0.2,k2v=(1,2,3)"'
'\b\n Refer to documentation for more information')
@click.option('--filter', 'filter_expr', default=None, show_default=True,
- help='\b Filter parameter points by expression.\n'
- '\b Example: "mass=2*,350,400,450;klambda=1.*,2.*,-1.*,-2.*"\n'
- '\b Refer to documentation for more information\n')
+ help='\b\n Filter parameter points by expression.'
+ '\b\n Example: "mass=(2*,350,400,450);klambda=(1.*,2.*,-1.*,-2.*)"'
+ '\b\n Refer to documentation for more information')
@click.option('--exclude', 'exclude_expr', default=None, show_default=True,
- help='\b Exclude parameter points by expression.\n'
- '\b Example: "mass=2*,350,400,450;klambda=1.*,2.*,-1.*,-2.*"\n'
- '\b Refer to documentation for more information\n')
+ help='\b\n Exclude parameter points by expression.'
+ '\b\n Example: "mass=(2*,350,400,450);klambda=(1.*,2.*,-1.*,-2.*)"'
+ '\b\n Refer to documentation for more information')
@click.option('--outdir', default='output', show_default=True,
- help='Output directory where cached limit files and merged limit file are saved.')
+ help='Output directory where cached limit files and the merged limit file are saved.')
@click.option('--cachedir', default='cache', show_default=True,
help='Cache directory relative to the output directory.')
@click.option('-o', '--outname', default='limits.json', show_default=True,
- help='Name of output limit file (all parameter points merged).')
+ help='Name of the output limit file (all parameter points merged).')
@click.option('--cache/--no-cache', default=True, show_default=True,
- help='Cache output of individual parameter point')
+ help='Cache output of individual parameter point.')
@click.option('--save-log/--no-log', default=True, show_default=True,
help='Save a log file for each parameter point.')
@click.option('--save-summary/--no-summary', default=True, show_default=False,
help='Save a summary file for each parameter point.')
-@click.option('--parallel', type=int, default=-1, show_default=True,
- help='\b\n Parallelize job across the N workers.'
- '\b\n Case 0: Jobs are run sequentially (for debugging).'
- '\b\n Case -1: Jobs are run across N_CPU workers.')
@click.option('-p', '--poi', 'poi_name', default=None,
help='Name of the parameter of interest (POI). If None, the first POI is used.')
@click.option('--mu_exp', type=float, default=0, show_default=True,
- help='Expected signal strengh value to be used for Asimov generation')
+ help='(DO NOT USE) Expected signal strengh value to be used for Asimov generation')
@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
- help='Name of dataset')
+ help='Name of dataset.')
@click.option('--asimov_data_name', 'asimov_data_name', default=None,
help='If given, use custom background asimov dataset instead of generating on the fly.')
@click.option('--blind/--unblind', 'do_blind', default=True, show_default=True,
- help='Blind/unblind analysis')
+ help='Blind/unblind analysis.')
@click.option('--CL', 'CL', type=float, default=0.95, show_default=True,
- help='CL value to use')
+ help='CL value to use.')
@click.option('--precision', default=0.005, show_default=True,
- help='precision in mu that defines iterative cutoff')
+ help='precision in mu that defines iterative cutoff.')
@click.option('--adjust_fit_range/--keep_fit_range', default=True, show_default=True,
- help='whether to adjust the fit range to median limit +- 5 sigma for observed fit')
+ help='whether to adjust the fit range to median limit +- 5 sigma for observed fit.')
@click.option('--do_tilde/--no_tilde', default=True, show_default=True,
- help='bound mu at zero if true and do the \tilde{q}_{mu} asymptotics')
+ help='bound mu at zero if true and do the \tilde{q}_{mu} asymptotics.')
@click.option('--predictive_fit/--no_predictive_fit', default=False, show_default=True,
- help='extrapolate best fit nuisance parameters based on previous fit results')
+ help='extrapolate best fit nuisance parameters based on previous fit results.')
@click.option('--do_better_bands/--skip_better_bands', default=True, show_default=True,
- help='evaluate asymptotic CLs limit for various sigma bands')
+ help='evaluate asymptotic CLs limit for various sigma bands.')
@click.option('--better_negative_bands/--skip_better_negative_bands', default=False, show_default=True,
- help='evaluate asymptotic CLs limit for negative sigma bands')
+ help='evaluate asymptotic CLs limit for negative sigma bands.')
@click.option('--binned/--unbinned', 'binned_likelihood', default=True, show_default=True,
- help='Binned likelihood')
+ help='Activate binned likelihood for RooRealSumPdf.')
@click.option('--save_log/--skip_log', default=True, show_default=True,
- help='Save log file')
+ help='Save log file.')
@click.option('--save_summary/--skip_summary', default=True, show_default=True,
- help='Save summary information')
+ help='Save summary information.')
@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
- help='Parameters to fix')
+ help='Parameters to fix.')
@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
- help='Parameters to profile')
+ help='Parameters to profile.')
@click.option('-w', '--workspace', 'ws_name', default=None, show_default=True,
help='Name of workspace. Auto-detect by default.')
@click.option('-m', '--model_config', 'mc_name', default=None, show_default=True,
help='Name of model config. Auto-detect by default.')
@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
- help='Name of initial snapshot')
+ help='Name of initial snapshot.')
+@click.option('--constrain/--no-constrain', 'constrain_nuis', default=True, show_default=True,
+ help='Use constrained NLL.')
@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
- help='Minimizer type')
+ help='Minimizer type.')
@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
- help='Minimizer algorithm')
+ help='Minimizer algorithm.')
+@click.option('--strategy', type=int, default=1, show_default=True,
+ help='Default minimization strategy.')
@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
- help='Convergence criterium')
+ help='Minimization convergence criterium.')
@click.option('--retry', type=int, default=2, show_default=True,
- help='Maximum number of retries upon a failed fit')
-@click.option('--strategy', type=int, default=1, show_default=True,
- help='Default minimization strategy')
-@click.option('--print_level', type=int, default=-1, show_default=True,
- help='Minimizer print level')
-@click.option('--timer/--no_timer', default=False, show_default=True,
- help='Enable minimizer timer')
-@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
- help='Number of CPUs to use per parameter')
-@click.option('--offset/--no-offset', default=True, show_default=True,
- help='Offset likelihood')
+ help='Maximum number of retries upon a failed fit.')
@click.option('--optimize', type=int, default=2, show_default=True,
- help='Optimize constant terms')
-@click.option('--improve', type=int, default=0, show_default=True,
- help='Execute improve after each minimization')
+ help='Optimize constant terms.')
@click.option('--minimizer_offset', type=int, default=1, show_default=True,
- help='Enable minimizer offsetting')
-@click.option('--fix-cache/--no-fix-cache', default=True, show_default=True,
- help='Fix StarMomentMorph cache')
-@click.option('--fix-multi/--no-fix-cache', default=True, show_default=True,
- help='Fix MultiPdf level 2')
-@click.option('--max_calls', type=int, default=-1, show_default=True,
- help='Maximum number of function calls')
-@click.option('--max_iters', type=int, default=-1, show_default=True,
- help='Maximum number of Minuit iterations')
+ help='Enable minimizer offsetting.')
+@click.option('--offset/--no-offset', default=True, show_default=True,
+ help='Offset likelihood.')
+@click.option('--binned/--unbinned', 'binned_likelihood', default=True, show_default=True,
+ help='Activate binned likelihood for RooRealSumPdf.')
+@click.option('--print_level', type=int, default=-1, show_default=True,
+ help='Minimizer print level.')
+@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
+ help='Number of CPUs to use during minimization.')
@click.option('--batch_mode/--no-batch', default=False, show_default=True,
- help='Batch mode when evaluating likelihood')
+ help='Batch mode when evaluating likelihood.')
@click.option('--int_bin_precision', type=float, default=-1., show_default=True,
help='Integrate the PDF over the bins instead of using the probability '
- 'density at the bin centre')
-@click.option('--constrain/--no-constrain', 'constrain_nuis', default=True, show_default=True,
- help='Use constrained NLL')
-@click.option('-v', '--verbosity', default="INFO", show_default=True,
- help='verbosity level ("DEBUG", "INFO", "WARNING", "ERROR")')
+ 'density at the bin center.')
+@click.option('--extra_minimizer_options', default=None, show_default=True,
+ help='Additional minimizer options to include. Format should be <config>=<value> '
+ 'separated by commas. Example: "discrete_min_tol=0.001,do_discrete_iteration=1"')
+@click.option('--cms_runtimedef', 'runtimedef_expr', default=None, show_default=True,
+ help='CMS specific runtime definitions. Format should be <config>=<value> '
+ 'separated by commas. Example: "REMOVE_CONSTANT_ZERO_POINT=1,ADDNLL_GAUSSNLL=0"')
+@click.option('--parallel', type=int, default=-1, show_default=True,
+ help='\b\n Parallelize job across the N workers.'
+ '\b\n Case 0: Jobs are run sequentially (for debugging).'
+ '\b\n Case -1: Jobs are run across N_CPU workers.')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
def limit_scan(**kwargs):
"""
Evaluate a set of parmeterised asymptotic cls limits
diff --git a/quickstats/clis/nuisance_parameter_tools.py b/quickstats/clis/nuisance_parameter_tools.py
new file mode 100644
index 0000000000000000000000000000000000000000..a85496cf68204ebd2604d4708e97eab0c7a9d50b
--- /dev/null
+++ b/quickstats/clis/nuisance_parameter_tools.py
@@ -0,0 +1,320 @@
+import os
+import json
+import click
+
+from .core import cli
+
+__all__ = ['harmonize_np', 'run_pulls', 'plot_pulls', 'np_correlation']
+
+@cli.command(name='harmonize_np')
+@click.argument('ws_files', nargs=-1)
+@click.option('-r', '--reference', required=True, help='Path to reference json file containing renaming scheme.')
+@click.option('-i', '--input_config_path', default=None, show_default=True,
+ help='Path to json file containing input workspace paths.')
+@click.option('-b', '--base_path', default='./', show_default=True,
+ help='Base path for input config.')
+@click.option('-o', '--outfile', default='renamed_np.json', show_default=True,
+ help='Output filename.')
+def harmonize_np(ws_files, reference, input_config_path, base_path, outfile):
+ """
+ Harmonize NP names across different workspaces
+ """
+ from quickstats.components import NuisanceParameterHarmonizer
+ harmonizer = NuisanceParameterHarmonizer(reference)
+ if (len(ws_files) > 0) and input_config_path is not None:
+ raise RuntimeError('either workspace paths or json file containing workspace paths should be given')
+ if len(ws_files) > 0:
+ harmonizer.harmonize(ws_files, outfile=outfile)
+ elif (input_config_path is not None):
+ harmonizer.harmonize_multi_input(input_config_path, base_path, outfile=outfile)
+
+@cli.command(name='run_pulls')
+@click.option('-i', '--input_file', 'filename', required=True,
+ help='Path to the input workspace file.')
+@click.option('-x', '--poi', 'poi_name', default=None,
+ help='POI to measure NP impact on.')
+@click.option('-o', '--outdir', default="pulls", show_default=True,
+ help='Output directory.')
+@click.option('-w', '--workspace', 'ws_name', default=None,
+ help='Name of workspace. Auto-detect by default.')
+@click.option('-m', '--model_config', 'mc_name', default=None,
+ help='Name of model config. Auto-detect by default.')
+@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
+ help='Name of dataset.')
+@click.option('--filter', 'filter_expr', default=None, show_default=True,
+ help='Filter nuisance parameter(s) to run pulls and impacts on.'+\
+ 'Multiple parameters are separated by commas.'+\
+ 'Wildcards are accepted. All NPs are included by default.')
+@click.option('-r', '--profile', 'profile_param', default=None, show_default=True,
+ help='Parameters to profile.')
+@click.option('-f', '--fix', 'fix_param', default=None, show_default=True,
+ help='Parameters to fix.')
+@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
+ help='Name of initial snapshot.')
+@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
+ help='Minimizer type.')
+@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
+ help='Minimizer algorithm.')
+@click.option('--strategy', type=int, default=1, show_default=True,
+ help='Default minimization strategy.')
+@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
+ help='Minimization convergence criterium.')
+@click.option('-q', '--precision', type=float, default=0.001, show_default=True,
+ help='Precision of sigma scan.')
+@click.option('--retry', type=int, default=1, show_default=True,
+ help='Maximum number of retries upon a failed fit.')
+@click.option('--optimize', type=int, default=2, show_default=True,
+ help='Optimize constant terms.')
+@click.option('--minimizer_offset', type=int, default=1, show_default=True,
+ help='Enable minimizer offsetting.')
+@click.option('--offset/--no-offset', default=True, show_default=True,
+ help='Offset likelihood.')
+@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
+ help='Number of CPUs to use during minimization.')
+@click.option('--print_level', type=int, default=-1, show_default=True,
+ help='Minimizer print level.')
+@click.option('--batch_mode/--no-batch', default=False, show_default=True,
+ help='Batch mode when evaluating likelihood.')
+@click.option('--int_bin_precision', type=float, default=-1., show_default=True,
+ help='Integrate the PDF over the bins instead of using the probability '
+ 'density at the bin center.')
+@click.option('--parallel', type=int, default=-1, show_default=True,
+ help='\b\n Parallelize job across the N workers.'
+ '\b\n Case 0: Jobs are run sequentially (for debugging).'
+ '\b\n Case -1: Jobs are run across N_CPU workers.')
+@click.option('--cache/--no-cache', default=True, show_default=True,
+ help='Cache existing result.')
+@click.option('--exclude', 'exclude_expr', default=None, show_default=True,
+ help='Exclude NPs to run pulls and impacts on. '+\
+ 'Multiple parameters are separated by commas.'+\
+ 'Wildcards are accepted.')
+@click.option('--save_log/--skip_log', default=True, show_default=True,
+ help='Save log file.')
+@click.option('--constrained_only/--any_nuis', default=True, show_default=True,
+ help='Whether to include constrained nuisance parameters only.')
+@click.option('--extra_minimizer_options', default=None, show_default=True,
+ help='Additional minimizer options to include. Format should be <config>=<value> '
+ 'separated by commas. Example: "discrete_min_tol=0.001,do_discrete_iteration=1"')
+@click.option('--cms_runtimedef', 'runtimedef_expr', default=None, show_default=True,
+ help='CMS specific runtime definitions. Format should be <config>=<value> '
+ 'separated by commas. Example: "REMOVE_CONSTANT_ZERO_POINT=1,ADDNLL_GAUSSNLL=0"')
+@click.option('--version', type=click.Choice(['1', '2']), default='2', show_default=True,
+ help='Version of tool to use (Choose between 1 and 2).')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
+def run_pulls(**kwargs):
+ """
+ Tool for computing NP pulls and impacts
+ """
+ version = kwargs.pop('version')
+ if version == '1':
+ from quickstats.components import NuisanceParameterPull
+ NuisanceParameterPull().run_pulls(**kwargs)
+ elif version == '2':
+ from quickstats.concurrent import NuisanceParameterRankingRunner
+ init_kwargs = {}
+ for key in ['filename', 'filter_expr', 'exclude_expr', 'poi_name',
+ 'data_name', 'cache', 'outdir', 'constrained_only',
+ 'save_log', 'parallel', 'verbosity']:
+ init_kwargs[key] = kwargs.pop(key)
+ init_kwargs['config'] = kwargs
+ runner = NuisanceParameterRankingRunner(**init_kwargs)
+ runner.run()
+
+@cli.command(name='plot_pulls')
+@click.option('-i', '--inputdir', required=True, help='Path to directory containing pull results')
+@click.option('-p', '--poi', default=None, help='Parameter of interest for plotting impact')
+@click.option('-n', '--n_rank', type=int, default=None, help='Total number of NP to rank')
+@click.option('-m', '--rank_per_plot', type=int, default=20, show_default=True,
+ help='Number of NP to show in a single plot.')
+@click.option('--ranking/--no_ranking', default=True, show_default=True,
+ help='Rank NP by impact.')
+@click.option('--threshold', type=float, default=0., show_default=True,
+ help='Filter NP by postfit impact threshold.')
+@click.option('--show_sigma/--hide_sigma', default=True, show_default=True,
+ help='Show one standard deviation pull.')
+@click.option('--show_prefit/--hide_prefit', default=True, show_default=True,
+ help='Show prefit impact.')
+@click.option('--show_postfit/--hide_postfit', default=True, show_default=True,
+ help='Show postfit impact.')
+@click.option('--sigma_bands/--no_sigma_bands', default=False, show_default=True,
+ help='Draw +-1, +-2 sigma bands.')
+@click.option('--sigma_lines/--no_sigma_lines', default=True, show_default=True,
+ help='Draw +-1 sigma lines.')
+@click.option('--ranking_label/--no_ranking_label', default=True, show_default=True,
+ help='Show ranking labels.')
+@click.option('--shade/--no_shade', default=True, show_default=True,
+ help='Draw shade.')
+@click.option('--correlation/--no_correlation', default=True, show_default=True,
+ help='Show correlation impact.')
+@click.option('--onesided/--overlap', default=True, show_default=True,
+ help='Show onesided impact.')
+@click.option('--relative/--absolute', default=False, show_default=True,
+ help='Show relative variation.')
+@click.option('--theta_max', type=float, default=2, show_default=True,
+ help='Pull range.')
+@click.option('-y', '--padding', type=int, default=7, show_default=True,
+ help='Padding below plot for texts and legends. NP column height is 1 unit.')
+@click.option('-h', '--height', type=float, default=1.0, show_default=True,
+ help='NP column height.')
+@click.option('-s', '--spacing', type=float, default=0., show_default=True,
+ help='Spacing between impact box.')
+@click.option('--label-fontsize', type=float, default=20., show_default=True,
+ help='Fontsize of analysis label text.')
+@click.option('-d', '--display_poi', default=r"$\mu$", show_default=True,
+ help='POI name to be shown in the plot.')
+@click.option('-t', '--extra_text', default=None, help='Extra texts below the ATLAS label. '+\
+ 'Use "//" as newline delimiter.')
+@click.option('--elumi_label/--no_elumi_label', default=True, show_default=True,
+ help='Show energy and luminosity labels.')
+@click.option('--ranking_label/--no_ranking_label', default=True, show_default=True,
+ help='Show ranking label.')
+@click.option('--energy', default="13 TeV", show_default=True,
+ help='Beam energy.')
+@click.option('--lumi', default="140 fb$^{-1}$", show_default=True,
+ help='Luminosity.')
+@click.option('--status', default="int", show_default=True,
+ help='\b\n Analysis status. Choose from'
+ '\b\n int : Internal'
+ '\b\n wip : Work in Progress'
+ '\b\n prelim : Preliminary'
+ '\b\n final : *no status label*'
+ '\b\n *custom input* : *custom input*')
+@click.option('--combine_pdf/--split_pdf', default=True, show_default=True,
+ help='Combine all ranking plots into a single pdf.')
+@click.option('--outdir', default='ranking_plots', show_default=True,
+ help='Output directory.')
+@click.option('-o', '--outname', default='ranking', show_default=True,
+ help='Output file name prefix.')
+@click.option('--style', default='default', show_default=True,
+ help='Plotting style. Built-in styles are "default" and "trex".'+\
+ 'Specify path to yaml file to set custom plotting style.')
+@click.option('--fix_axis_scale/--free_axis_scale', default=True, show_default=True,
+ help='Fix the axis scale across all ranking plots.')
+@click.option('--version', type=click.Choice(['1', '2']), default='2', show_default=True,
+ help='Version of tool to use (Choose between 1 and 2).')
+def plot_pulls(**kwargs):
+ """
+ Tool for plotting NP pulls and impact rankings
+ """
+ from quickstats.plots.np_ranking_plot import NPRankingPlot
+ inputdir, poi = kwargs.pop('inputdir'), kwargs.pop('poi')
+ version = kwargs.pop('version')
+ ranking_plot = NPRankingPlot(inputdir, poi, version=version)
+ ranking_plot.plot(**kwargs)
+
+@cli.command(name='np_correlation')
+@click.option('-i', '--input_file', "filename", required=True,
+ help='Path to the input workspace file.')
+@click.option('-o', '--basename', default='NP_correlation_matrix', show_default=True,
+ help='Base name of the output.')
+@click.option('--select', default=None, show_default=True,
+ help='Select specific NPs to be stored in the final output (for json and plot only). '
+ 'Use comma to separate the selection (wild card is supported).')
+@click.option('--remove', default=None, show_default=True,
+ help='Select specific NPs to be removed in the final output (for json and plot only). '
+ 'Use comma to separate the selection (wild card is supported).')
+@click.option('--save_plot/--no_save_plot', default=True, show_default=True,
+ help='Save NP correlation matrix as a plot in pdf format')
+@click.option('--save_json/--no_save_json', default=False, show_default=True,
+ help='Save NP correlation matrix as a json file')
+@click.option('--save_root/--no_save_root', default=False, show_default=True,
+ help='Save NP correlation matrix as a 2D histogram in a root file')
+@click.option('--plot_style', default="default", show_default=True,
+ help='Plot style if save_plot is enabled. Choose between \"default\" and '
+ f'\"viridis\". Alternatively, a path to a yaml config file can be used')
+@click.option('-w', '--workspace', 'ws_name', default=None, show_default=True,
+ help='Name of workspace. Auto-detect by default.')
+@click.option('-m', '--model_config', 'mc_name', default=None, show_default=True,
+ help='Name of model config. Auto-detect by default.')
+@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
+ help='Name of dataset.')
+@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
+ help='Name of initial snapshot.')
+@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
+ help='Parameters to profile.')
+@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
+ help='Parameters to fix.')
+@click.option('--constrain/--no-constrain', 'constrain_nuis', default=True, show_default=True,
+ help='Use constrained NLL (i.e. include systematics)')
+@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
+ help='Minimizer type.')
+@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
+ help='Minimizer algorithm.')
+@click.option('--strategy', type=int, default=1, show_default=True,
+ help='Default minimization strategy.')
+@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
+ help='Minimization convergence criterium.')
+@click.option('--retry', type=int, default=1, show_default=True,
+ help='Maximum number of retries upon a failed fit.')
+@click.option('--optimize', type=int, default=2, show_default=True,
+ help='Optimize constant terms.')
+@click.option('--minimizer_offset', type=int, default=1, show_default=True,
+ help='Enable minimizer offsetting.')
+@click.option('--offset/--no-offset', default=True, show_default=True,
+ help='Offset likelihood.')
+@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
+ help='Number of CPUs to use during minimization.')
+@click.option('--batch_mode/--no-batch', default=False, show_default=True,
+ help='Batch mode when evaluating likelihood.')
+@click.option('--int_bin_precision', type=float, default=-1., show_default=True,
+ help='Integrate the PDF over the bins instead of using the probability '
+ 'density at the bin center.')
+@click.option('--extra_minimizer_options', default=None, show_default=True,
+ help='Additional minimizer options to include. Format should be <config>=<value> '
+ 'separated by commas. Example: "discrete_min_tol=0.001,do_discrete_iteration=1"')
+@click.option('--cms_runtimedef', 'runtimedef_expr', default=None, show_default=True,
+ help='CMS specific runtime definitions. Format should be <config>=<value> '
+ 'separated by commas. Example: "REMOVE_CONSTANT_ZERO_POINT=1,ADDNLL_GAUSSNLL=0"')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
+def np_correlation(**kwargs):
+ """
+ Evaluate post-fit NP correlation matrix
+ """
+ _kwargs = {}
+ for arg_name in ["basename", "save_plot", "save_json", "save_root",
+ "plot_style", "select", "remove"]:
+ _kwargs[arg_name] = kwargs.pop(arg_name)
+ _init_kwargs = {}
+ for arg_name in ["filename", "data_name", "verbosity"]:
+ _init_kwargs[arg_name] = kwargs.pop(arg_name)
+ _init_kwargs['config'] = kwargs
+ _init_kwargs['poi_name'] = []
+ from quickstats.components import AnalysisBase
+ analysis = AnalysisBase(**_init_kwargs)
+ analysis.minimizer.configure(hesse=True)
+ analysis.nll_fit(mode=3)
+ fit_result = analysis.roofit_result
+ basename = os.path.splitext(_kwargs['basename'])[0]
+ from quickstats.utils.roofit_utils import get_correlation_matrix
+ if _kwargs['save_root']:
+ correlation_hist = get_correlation_matrix(fit_result, lib="root")
+ outname = basename + ".root"
+ correlation_hist.SaveAs(outname)
+ print(f"INFO: Saved correlation histogram to `{outname}`")
+ correlation_hist.Delete()
+ from quickstats.utils.common_utils import filter_by_wildcards
+ if _kwargs['save_json'] or _kwargs['save_plot']:
+ df = get_correlation_matrix(fit_result, lib='pandas')
+ labels = list(df.columns)
+ selected = filter_by_wildcards(labels, _kwargs['select'])
+ selected = filter_by_wildcards(selected, _kwargs['remove'], exclusion=True)
+ to_drop = list(set(labels) - set(selected))
+ df = df.drop(to_drop, axis=0).drop(to_drop, axis=1).transpose()
+ if _kwargs['save_json']:
+ data = df.to_dict()
+ outname = basename + ".json"
+ with open(outname, "w") as out:
+ json.dump(data, out, indent=2)
+ print(f"INFO: Saved correlation data to `{outname}`")
+ if _kwargs['save_plot']:
+ import matplotlib.pyplot as plt
+ from quickstats.plots import CorrelationPlot
+ plotter = CorrelationPlot(df)
+ ax = plotter.draw_style(_kwargs['plot_style'])
+ outname = basename + ".pdf"
+ plt.savefig(outname, bbox_inches="tight")
+ print(f"INFO: Saved correlation plot to `{outname}`")
\ No newline at end of file
diff --git a/quickstats/clis/processor_cli.py b/quickstats/clis/processor_tools.py
similarity index 77%
rename from quickstats/clis/processor_cli.py
rename to quickstats/clis/processor_tools.py
index c4994cd65fc6c6edc4626a4008783681c420de0a..97870a3b031127e4f45b1be0010732bdff381fa7 100644
--- a/quickstats/clis/processor_cli.py
+++ b/quickstats/clis/processor_tools.py
@@ -1,7 +1,11 @@
import os
import click
-@click.command(name='process_rfile')
+from .core import cli
+
+__all__ = ['process_rfile']
+
+@cli.command(name='process_rfile')
@click.option('-i', '--input_file', 'filename', required=True,
help='Input ROOT file to process.')
@click.option('-c', '--config', 'config_path', required=True,
@@ -9,14 +13,15 @@ import click
@click.option('--multithread/--no-multirhread', default=True, show_default=True,
help='Enable implicit multi-threading.')
@click.option('-g', '--global', 'glob', default=None,
- help='Include global variables in the form "<name>=<value>,..."')
+ help='Include global variables in the form "<name>=<value>,..." .')
@click.option('-f', '--flag', default=None,
help='Flags to set (separated by commas).')
-@click.option('-v', '--verbosity', default="INFO", show_default=True,
- help='verbosity level ("DEBUG", "INFO", "WARNING", "ERROR")')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
def process_rfile(filename, config_path, multithread, glob, flag, verbosity):
"""
- Process a ROOT file based on RDataFrame routines.
+ Process a ROOT file based on RDataFrame routines
"""
from quickstats.components.processors import RooProcessor
from quickstats.components.processors.actions import RooProcGlobalVariables
diff --git a/quickstats/clis/stat_tools.py b/quickstats/clis/stat_tools.py
new file mode 100644
index 0000000000000000000000000000000000000000..ef8b833da8b0e3ffe382353546663dfd798b0b10
--- /dev/null
+++ b/quickstats/clis/stat_tools.py
@@ -0,0 +1,201 @@
+import os
+import json
+import click
+
+from .core import cli
+
+__all__ = ['generate_standard_asimov', 'toy_significance', 'toy_limit']
+
+@cli.command(name='generate_standard_asimov')
+@click.option('-i', '--input_file', 'filename', required=True,
+ help='Path to the input workspace file.')
+@click.option('-o', '--output_file', 'outname', required=True,
+ help='Name of the output workspace containing the '
+ 'generated asimov dataset.')
+@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
+ help='Name of the dataset used in NP profiling.')
+@click.option('-p', '--poi', 'poi_name', required=True,
+ help='Name of the parameter of interest (POI). Multiple POIs are separated by commas.')
+@click.option('-s', '--poi_scale', type=float, default=1.0, show_default=True,
+ help='Scale factor applied to the poi value.')
+@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
+ help='Minimization convergence criterium.')
+@click.option('--strategy', type=int, default=1, show_default=True,
+ help='Default minimization strategy.')
+@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
+ help='Parameters to fix.')
+@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
+ help='Parameters to profile.')
+@click.option('-s', '--snapshot', 'snapshot_name', default=None, show_default=True,
+ help='Name of initial snapshot.')
+@click.option('--rebuild/--do-not-rebuild', default=False, show_default=True,
+ help='Rebuild the workspace.')
+@click.option('--asimov_names', default=None, show_default=True,
+ help='Names of the output asimov datasets (separated by commas). If not specified, '
+ 'a default name for the corresponding asimov type will be given.')
+@click.option('--asimov_snapshots', default=None, show_default=True,
+ help='Names of the output asimov snapshots (separated by commas). If not specified, '
+ 'a default name for the corresponding asimov type will be given.')
+@click.option('-t', '--asimov_types', default="0,1,2", show_default=True,
+ help='\b\n Types of asimov dataset to generate separated by commas.'
+ '\b\n 0: fit with POI fixed to 0'
+ '\b\n 1: fit with POI fixed to 1'
+ '\b\n 2: fit with POI free and set POI to 1 after fit'
+ '\b\n 3: fit with POI and constrained NP fixed to 0'
+ '\b\n 4: fit with POI fixed to 1 and constrained NP fixed to 0'
+ '\b\n 5: fit with POI free and constrained NP fixed to 0 and set POI to 1 after fit'
+ '\b\n -1: nominal NP with POI set to 0'
+ '\b\n -2: nominal NP with POI set to 1')
+@click.option('--extra_minimizer_options', default=None, show_default=True,
+ help='Additional minimizer options to include. Format should be <config>=<value> '
+ 'separated by commas. Example: "discrete_min_tol=0.001,do_discrete_iteration=1"')
+@click.option('--cms_runtimedef', 'runtimedef_expr', default=None, show_default=True,
+ help='CMS specific runtime definitions. Format should be <config>=<value> '
+ 'separated by commas. Example: "REMOVE_CONSTANT_ZERO_POINT=1,ADDNLL_GAUSSNLL=0"')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
+def generate_standard_asimov(**kwargs):
+ """
+ Generate standard Asimov dataset
+ """
+ from quickstats.components import AsimovGenerator
+ from quickstats.utils.string_utils import split_str
+ outname = kwargs.pop('outname')
+ asimov_types = kwargs.pop('asimov_types')
+ try:
+ asimov_types = split_str(asimov_types, sep=",", cast=int)
+ except:
+ asimov_types = split_str(asimov_types, sep=",")
+ fix_param = kwargs.pop('fix_param')
+ profile_param = kwargs.pop('profile_param')
+ snapshot_name = kwargs.pop('snapshot_name')
+ poi_scale = kwargs.pop("poi_scale")
+ asimov_names = kwargs.pop("asimov_names")
+ asimov_snapshots = kwargs.pop("asimov_snapshots")
+ verbosity = kwargs.pop("verbosity")
+ rebuild = kwargs.pop("rebuild")
+ kwargs['poi_name'] = split_str(kwargs.pop('poi_name'), sep=",")
+ config = {
+ 'fix_param': fix_param,
+ 'profile_param': profile_param,
+ 'snapshot_name': snapshot_name
+ }
+ from quickstats.utils.string_utils import split_str
+ if asimov_names is not None:
+ asimov_names = split_str(asimov_names, sep=",")
+ if asimov_snapshots is not None:
+ asimov_snapshots = split_str(asimov_snapshots, sep=",")
+ generator = AsimovGenerator(**kwargs, config=config, verbosity=verbosity)
+ generator.generate_standard_asimov(asimov_types, poi_scale=poi_scale,
+ asimov_names=asimov_names,
+ asimov_snapshots=asimov_snapshots)
+ generator.save(outname, rebuild=rebuild)
+
+@cli.command(name='toy_significance')
+@click.option('-i', '--input_file', 'filename', required=True,
+ help='Path to the input workspace file.')
+@click.option('-o', '--output_file', 'outname', default="toy_study/results.json",
+ help='Name of the output file containing toy results.')
+@click.option('-n', '--n_toys', type=int,
+ help='Number of the toys to use.')
+@click.option('-b', '--batchsize', type=int, default=100, show_default=True,
+ help='Divide the task into batches each containing this number of toys. '
+ 'Result from each batch is saved for caching and different batches '
+ 'are run in parallel if needed.')
+@click.option('-s', '--seed', type=int, default=0, show_default=True,
+ help='Random seed used for generating toy datasets.')
+@click.option('-p', '--poi', 'poi_name', default=None,
+ help='Name of the parameter of interest (POI). If None, the first POI is used.')
+@click.option('-v', '--poi_val', type=float, default=0, show_default=True,
+ help='POI value when generating the toy dataset.')
+@click.option('--binned/--unbinned', default=True, show_default=True,
+ help='Generate binned toy dataset.')
+@click.option('--cache/--no-cache', default=True, show_default=True,
+ help='Cache existing batch results.')
+@click.option('--fit_options', default=None, help='A json file specifying the fit options.')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
+@click.option('--parallel', type=int, default=-1, show_default=True,
+ help='\b\n Parallelize job across the N workers.'
+ '\b\n Case 0: Jobs are run sequentially (for debugging).'
+ '\b\n Case -1: Jobs are run across N_CPU workers.')
+def toy_significance(**kwargs):
+ """
+ Generate toys and evaluate significance
+ """
+ from quickstats.components import PValueToys
+ n_toys = kwargs.pop("n_toys")
+ batchsize = kwargs.pop("batchsize")
+ seed = kwargs.pop("seed")
+ cache = kwargs.pop("cache")
+ outname = kwargs.pop("outname")
+ parallel = kwargs.pop("parallel")
+ pvalue_toys = PValueToys(**kwargs)
+ pvalue_toys.get_toy_results(n_toys=n_toys, batchsize=batchsize, seed=seed,
+ cache=cache, save_as=outname, parallel=parallel)
+
+@cli.command(name='toy_limit')
+@click.option('-i', '--input_file', 'filename', required=True,
+ help='Path to the input workspace file.')
+@click.option('-d', '--data', 'data_name', default='combData', show_default=True,
+ help='Name of the dataset used for computing observed limit.')
+@click.option('-o', '--output_file', 'outname',
+ default="toy_study/toy_result_seed_{seed}_batch_{batch}.root",
+ show_default=True,
+ help='Name of the output file containing toy results.')
+@click.option('--poi_max', type=float, default=None,
+ help='Maximum range of POI.')
+@click.option('--poi_min', type=float, default=None,
+ help='Minimum range of POI.')
+@click.option('--scan_max', type=float, default=None,
+ help='Maximum scan value of POI.')
+@click.option('--scan_min', type=float, default=None,
+ help='Minimum scan value of POI.')
+@click.option('--steps', type=int, default=10, show_default=True,
+ help='Number of scan steps.')
+@click.option('--mu_val', type=float, default=None,
+ help='Value of POI for running a single point.')
+@click.option('-n', '--n_toys', type=int,
+ help='Number of the toys to use.')
+@click.option('-b', '--batchsize', type=int, default=50, show_default=True,
+ help='Divide the task into batches each containing this number of toys. '
+ 'Result from each batch is saved for caching and different batches '
+ 'are run in parallel if needed.')
+@click.option('-s', '--seed', type=int, default=2021, show_default=True,
+ help='Random seed used for generating toy datasets.')
+@click.option('-t', '--tolerance', type=float, default=1., show_default=True,
+ help='Tolerance for minimization.')
+@click.option('-p', '--poi', 'poi_name', default=None,
+ help='Name of the parameter of interest (POI). If None, the first POI is used.')
+@click.option('--minimizer_type', default="Minuit2", show_default=True,
+ help='Minimizer type.')
+@click.option('--strategy', type=int, default=1, show_default=True,
+ help='Default minimization strategy.')
+@click.option('--offset/--no-offset', default=True, show_default=True,
+ help='Use NLL offset.')
+@click.option('--print_level', type=int, default=-1, show_default=True,
+ help='Minimizer print level.')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
+@click.option('-f', '--fix', 'fix_param', default="", show_default=True,
+ help='Parameters to fix.')
+@click.option('-r', '--profile', 'profile_param', default="", show_default=True,
+ help='Parameters to profile.')
+@click.option('--snapshot', 'snapshot_name', default=None, help='Name of initial snapshot')
+@click.option('--parallel', type=int, default=-1, show_default=True,
+ help='\b\n Parallelize job across the N workers.'
+ '\b\n Case 0: Jobs are run sequentially (for debugging).'
+ '\b\n Case -1: Jobs are run across N_CPU workers.')
+def toy_limit(**kwargs):
+ """
+ Generate toys and evaluate limits
+ """
+ from quickstats.components.toy_limit_calculator import evaluate_batched_toy_limits
+ if not (((kwargs['scan_min'] is None) and (kwargs['scan_max'] is None) and (kwargs['mu_val'] is not None)) or \
+ ((kwargs['scan_min'] is not None) and (kwargs['scan_max'] is not None) and (kwargs['mu_val'] is None))):
+ raise ValueError("please provide either (scan_min, scan_max, steps) for running a scan or (mu_val)"
+ " for running a single point")
+ evaluate_batched_toy_limits(**kwargs)
\ No newline at end of file
diff --git a/quickstats/clis/workspace_tools.py b/quickstats/clis/workspace_tools.py
index e5ffe18f74646b5272a50e557d321e9b75d29516..47d8be470e9aa59e3d29a60fe645eefdd3542d18 100644
--- a/quickstats/clis/workspace_tools.py
+++ b/quickstats/clis/workspace_tools.py
@@ -1,13 +1,49 @@
import os
import click
-from .core import DelimitedStr
+from .core import DelimitedStr, cli
-@click.command(name='build_xml_ws')
+__all__ = ['inspect_ws', 'compare_ws', 'build_xml_ws', 'modify_ws', 'combine_ws', 'decompose_ws']
+
+kItemChoices = ['workspace', 'dataset', 'snapshot', 'category', 'poi',
+ 'detailed_nuisance_parameter', 'nuisance_parameter',
+ 'global_observable', 'auxiliary']
+kDefaultItems = ",".join(['workspace', 'dataset', 'snapshot', 'category',
+ 'poi', 'detailed_nuisance_parameter'])
+
+@cli.command(name='inspect_ws')
+@click.option('-i', '--input_file', required=True, help='Path to the input workspace file.')
+@click.option('-w', '--workspace', 'ws_name', default=None, help='Name of workspace. Auto-detect by default.')
+@click.option('-d', '--dataset', 'data_name', default=None, help='Name of dataset. Generally not needed.')
+@click.option('-m', '--model_config', 'mc_name', default=None, help='Name of model config. Auto-detect by default.')
+@click.option('-o', '--output_file', default=None, help='Export output to a text file. If None, no output is saved.')
+@click.option('--items', cls=DelimitedStr, type=click.Choice(kItemChoices), show_default=True,
+ default=kDefaultItems, help='Items to include in the summary (separated by commas).')
+@click.option('--include', 'include_patterns', default=None,
+ help='Match variable names with given patterns (separated by commas).')
+@click.option('--exclude', 'exclude_patterns', default=None,
+ help='Exclude variable names with given patterns (separated by commas).')
+@click.option('--detailed/--name-only', default=True, show_default=True,
+ help='Include detailed variable properties or just the variable name in the summary.')
+def inspect_ws(input_file, ws_name=None, data_name=None, mc_name=None, output_file=None, items=None,
+ include_patterns=None, exclude_patterns=None, detailed=True):
+ '''
+ Inspect workspace attributes
+ '''
+ from quickstats.components import ExtendedModel
+ model = ExtendedModel(input_file, ws_name=ws_name, mc_name=mc_name, data_name=data_name,
+ verbosity="WARNING")
+ from quickstats.utils.string_utils import split_str
+ #items = items.split(",") if items is not None else None
+ include_patterns = split_str(include_patterns, ',') if include_patterns is not None else None
+ exclude_patterns = split_str(exclude_patterns, ',') if exclude_patterns is not None else None
+ model.stdout.verbosity = "INFO"
+ model.print_summary(items=items, save_as=output_file, detailed=detailed,
+ include_patterns=include_patterns, exclude_patterns=exclude_patterns)
+
+@cli.command(name='build_xml_ws')
@click.option('-i', '--filename', 'source', required=True,
help='Input xml file.')
-@click.option('--binned/--unbinned', 'use_binned', default=False, show_default=True,
- help='Fit to binned data.')
@click.option('--data_storage_type', default="vector", show_default=True,
type=click.Choice(['vector', 'tree', 'composite']),
help='Set RooAbsData StorageType. Available choices: "vector", "tree", "composite".')
@@ -15,35 +51,34 @@ from .core import DelimitedStr
help='Base directory to which files in the xmls are referenced. '
'By default, the directory of the input xml file is used.')
@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
- help='Minimizer type')
+ help='Minimizer type.')
@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
- help='Minimizer algorithm')
-@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
- help='Number of CPUs to use per parameter')
+ help='Minimizer algorithm.')
+@click.option('--strategy', type=int, default=1, show_default=True,
+ help='Default minimization strategy.')
@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
- help='Convergence criterium')
+ help='Minimization convergence criterium.')
@click.option('--retry', type=int, default=1, show_default=True,
- help='Maximum number of retries upon a failed fit')
-@click.option('--strategy', type=int, default=1, show_default=True,
- help='Default minimization strategy')
-@click.option('--print_level', type=int, default=-1, show_default=True,
- help='Minimizer print level')
-@click.option('--fix-cache/--no-fix-cache', default=True, show_default=True,
- help='Fix StarMomentMorph cache')
-@click.option('--fix-multi/--no-fix-cache', default=True, show_default=True,
- help='Fix MultiPdf level 2')
-@click.option('--max_calls', type=int, default=-1, show_default=True,
- help='Maximum number of function calls')
-@click.option('--max_iters', type=int, default=-1, show_default=True,
- help='Maximum number of Minuit iterations')
+ help='Maximum number of retries upon a failed fit.')
@click.option('--optimize', type=int, default=2, show_default=True,
- help='Optimize constant terms')
+ help='Optimize constant terms.')
+@click.option('--minimizer_offset', type=int, default=1, show_default=True,
+ help='Enable minimizer offsetting.')
@click.option('--offset/--no-offset', default=True, show_default=True,
- help='Offset likelihood')
+ help='Offset likelihood.')
+@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
+ help='Number of CPUs to use during minimization.')
@click.option('--apply-fix/--do-not-apply-fix', default=False, show_default=True,
- help='Apply a fix on the up/down uncertainty implementation')
-@click.option('-v', '--verbosity', default="INFO", show_default=True,
- help='Verbosity level ("DEBUG", "INFO", "WARNING", "ERROR")')
+ help='Apply a fix on the up/down uncertainty implementation.')
+@click.option('--extra_minimizer_options', default=None, show_default=True,
+ help='Additional minimizer options to include. Format should be <config>=<value> '
+ 'separated by commas. Example: "discrete_min_tol=0.001,do_discrete_iteration=1"')
+@click.option('--cms_runtimedef', 'runtimedef_expr', default=None, show_default=True,
+ help='CMS specific runtime definitions. Format should be <config>=<value> '
+ 'separated by commas. Example: "REMOVE_CONSTANT_ZERO_POINT=1,ADDNLL_GAUSSNLL=0"')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
@click.option('--version', type=int, default=2, show_default=True,
help='Version of XMLWSBuilder to use (Choose between 1 and 2).')
def build_xml_ws(**kwargs):
@@ -63,7 +98,7 @@ def build_xml_ws(**kwargs):
builder = XMLWSBuilder(**_kwargs)
builder.generate_workspace()
-@click.command(name='modify_ws')
+@cli.command(name='modify_ws')
@click.option('-i', '--filename', 'source', required=True,
help='Input xml/json file.')
@click.option('--input_workspace',
@@ -72,35 +107,34 @@ def build_xml_ws(**kwargs):
help='Override output workspace path from the xml/json file.')
@click.option('--import-class-code/--no-import-class-code', 'import_class_code',
default=True, show_default=True,
- help='Import class code')
+ help='Import class code.')
@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
- help='Minimizer type')
+ help='Minimizer type.')
@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
- help='Minimizer algorithm')
-@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
- help='Number of CPUs to use per parameter')
-@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
- help='Convergence criterium')
-@click.option('--retry', type=int, default=0, show_default=True,
- help='Maximum number of retries upon a failed fit')
+ help='Minimizer algorithm.')
@click.option('--strategy', type=int, default=1, show_default=True,
- help='Default minimization strategy')
-@click.option('--print_level', type=int, default=-1, show_default=True,
- help='Minimizer print level')
-@click.option('--fix-cache/--no-fix-cache', default=True, show_default=True,
- help='Fix StarMomentMorph cache')
-@click.option('--fix-multi/--no-fix-cache', default=True, show_default=True,
- help='Fix MultiPdf level 2')
-@click.option('--max_calls', type=int, default=-1, show_default=True,
- help='Maximum number of function calls')
-@click.option('--max_iters', type=int, default=-1, show_default=True,
- help='Maximum number of Minuit iterations')
+ help='Default minimization strategy.')
+@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
+ help='Minimization convergence criterium.')
+@click.option('--retry', type=int, default=1, show_default=True,
+ help='Maximum number of retries upon a failed fit.')
@click.option('--optimize', type=int, default=2, show_default=True,
- help='Optimize constant terms')
+ help='Optimize constant terms.')
+@click.option('--minimizer_offset', type=int, default=1, show_default=True,
+ help='Enable minimizer offsetting.')
@click.option('--offset/--no-offset', default=True, show_default=True,
- help='Offset likelihood')
-@click.option('-v', '--verbosity', default="INFO", show_default=True,
- help='verbosity level ("DEBUG", "INFO", "WARNING", "ERROR")')
+ help='Offset likelihood.')
+@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
+ help='Number of CPUs to use during minimization.')
+@click.option('--extra_minimizer_options', default=None, show_default=True,
+ help='Additional minimizer options to include. Format should be <config>=<value> '
+ 'separated by commas. Example: "discrete_min_tol=0.001,do_discrete_iteration=1"')
+@click.option('--cms_runtimedef', 'runtimedef_expr', default=None, show_default=True,
+ help='CMS specific runtime definitions. Format should be <config>=<value> '
+ 'separated by commas. Example: "REMOVE_CONSTANT_ZERO_POINT=1,ADDNLL_GAUSSNLL=0"')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
def modify_ws(**kwargs):
"""
Modify workspace from XML/json config files
@@ -128,7 +162,7 @@ kDefaultVisibility = ",".join(["workspace=0b11001", "category=0b10011", "snapsho
"dataset=0b11011", "pdf=0b01011", "function=0b01011" , "poi=0b01011",
"nuisance_parameter=0b01011", "global_observable=0b01011", "auxiliary=0b01011"])
-@click.command(name='compare_ws')
+@cli.command(name='compare_ws')
@click.option('-l', '--left', required=True,
help='Path to the input workspace file (left of comparison).')
@click.option('-r', '--right', required=True,
@@ -153,8 +187,9 @@ kDefaultVisibility = ",".join(["workspace=0b11001", "category=0b10011", "snapsho
help='Save comparison data as a json file.')
@click.option('--save_excel_data', default=None,
help='Save comparison data as an excel file.')
-@click.option('-v', '--verbosity', default="INFO", show_default=True,
- help='verbosity level ("DEBUG", "INFO", "WARNING", "ERROR")')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
def compare_ws(**kwargs):
"""
Compare two workspace files
@@ -178,7 +213,7 @@ def compare_ws(**kwargs):
comparer.save_excel(save_excel_data)
-@click.command(name='combine_ws')
+@cli.command(name='combine_ws')
@click.option('-i', '--filename', 'source', required=True,
help='Input xml file.')
@click.option('--input_workspace',
@@ -186,40 +221,39 @@ def compare_ws(**kwargs):
@click.option('--output_workspace',
help='Override output workspace path from the xml file.')
@click.option('--save_rename_ws/--skip_rename_ws', default=False, show_default=True,
- help='Save a temporary workspace after the rename step')
+ help='Save a temporary workspace after the rename step.')
@click.option('--save_combine_ws/--skip_combine_ws', default=False, show_default=True,
- help='Save a temporary workspace after the combine step')
+ help='Save a temporary workspace after the combine step.')
@click.option('-t', '--minimizer_type', default="Minuit2", show_default=True,
- help='Minimizer type')
+ help='Minimizer type.')
@click.option('-a', '--minimizer_algo', default="Migrad", show_default=True,
- help='Minimizer algorithm')
-@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
- help='Number of CPUs to use per parameter')
-@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
- help='Convergence criterium')
-@click.option('--retry', type=int, default=0, show_default=True,
- help='Maximum number of retries upon a failed fit')
+ help='Minimizer algorithm.')
@click.option('--strategy', type=int, default=1, show_default=True,
- help='Default minimization strategy')
-@click.option('--print_level', type=int, default=-1, show_default=True,
- help='Minimizer print level')
-@click.option('--fix-cache/--no-fix-cache', default=True, show_default=True,
- help='Fix StarMomentMorph cache')
-@click.option('--fix-multi/--no-fix-cache', default=True, show_default=True,
- help='Fix MultiPdf level 2')
-@click.option('--max_calls', type=int, default=-1, show_default=True,
- help='Maximum number of function calls')
-@click.option('--max_iters', type=int, default=-1, show_default=True,
- help='Maximum number of Minuit iterations')
+ help='Default minimization strategy.')
+@click.option('-e', '--eps', type=float, default=1.0, show_default=True,
+ help='Minimization convergence criterium.')
+@click.option('--retry', type=int, default=1, show_default=True,
+ help='Maximum number of retries upon a failed fit.')
@click.option('--optimize', type=int, default=2, show_default=True,
- help='Optimize constant terms')
+ help='Optimize constant terms.')
+@click.option('--minimizer_offset', type=int, default=1, show_default=True,
+ help='Enable minimizer offsetting.')
@click.option('--offset/--no-offset', default=True, show_default=True,
- help='Offset likelihood')
+ help='Offset likelihood.')
+@click.option('-c', '--num_cpu', type=int, default=1, show_default=True,
+ help='Number of CPUs to use during minimization.')
@click.option('--import-class-code/--no-import-class-code', 'import_class_code',
default=True, show_default=True,
- help='Import class code')
-@click.option('-v', '--verbosity', default="INFO", show_default=True,
- help='verbosity level ("DEBUG", "INFO", "WARNING", "ERROR")')
+ help='Import class code.')
+@click.option('--extra_minimizer_options', default=None, show_default=True,
+ help='Additional minimizer options to include. Format should be <config>=<value> '
+ 'separated by commas. Example: "discrete_min_tol=0.001,do_discrete_iteration=1"')
+@click.option('--cms_runtimedef', 'runtimedef_expr', default=None, show_default=True,
+ help='CMS specific runtime definitions. Format should be <config>=<value> '
+ 'separated by commas. Example: "REMOVE_CONSTANT_ZERO_POINT=1,ADDNLL_GAUSSNLL=0"')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
def combine_ws(**kwargs):
"""
Combine workspace from XML config files
@@ -253,31 +287,32 @@ def combine_ws(**kwargs):
import_class_code=import_class_code)
-@click.command(name='decompose_ws')
+@cli.command(name='decompose_ws')
@click.option('-i', '--infile', required=True,
- help='Path to the input workspace file')
+ help='Path to the input workspace file.')
@click.option('-o', '--outfile', required=True,
- help='Path to the output workspace file')
+ help='Path to the output workspace file.')
@click.option('--import-class-code/--no-import-class-code', 'import_class_code',
default=True, show_default=True,
- help='Import class code')
+ help='Import class code.')
@click.option('-c', '--category_expr', default='*', show_default=True,
help='Categories to keep in the decomposed workspace (separated by commas). '
'Both category index and category label can be used. Category index can be '
'a single number or a range "<min_index>-<max_index>. Wildcard is supported '
'for category labels.')
-@click.option('-v', '--verbosity', default="INFO", show_default=True,
- help='verbosity level ("DEBUG", "INFO", "WARNING", "ERROR")')
@click.option('--snapshots', 'snapshots_to_save', default=None, show_default=True,
help='Snapshots to save (separated by commas)'
- 'By default, all existing snapshots will be saved')
+ 'By default, all existing snapshots will be saved.')
@click.option('--rebuild-nuis/--no-rebuild-nuis', default=False, show_default=True,
- help='Whether to rebuild the nuisance parameter set')
+ help='Whether to rebuild the nuisance parameter set.')
@click.option('--rebuild-pdf/--no-rebuild-pdf', default=False, show_default=True,
- help='Whether to rebuild category pdfs')
+ help='Whether to rebuild category pdfs.')
+@click.option('-v', '--verbosity', default='INFO', show_default=True,
+ type=click.Choice(["DEBUG", "INFO", "WARNING", "ERROR"], case_sensitive=False),
+ help='Verbosity level.')
def decompose_ws(**kwargs):
"""
- Decompose workspace into subcategories.
+ Decompose workspace into subcategories
"""
init_kwargs = {}
for key in ['verbosity']:
diff --git a/quickstats/components/__init__.py b/quickstats/components/__init__.py
index 9ad30f2c1219faeafcc756ca29be2961c6acff38..ba4e6c902be8601567efe8f1a0ce544bc98ca6bb 100644
--- a/quickstats/components/__init__.py
+++ b/quickstats/components/__init__.py
@@ -1,5 +1,6 @@
from .basics import *
from .root_object import ROOTObject
+from .discrete_nuisance import DiscreteNuisance
from .extended_minimizer import ExtendedMinimizer
from .extended_model import ExtendedModel
from .analysis_object import AnalysisObject
@@ -15,6 +16,7 @@ from .extended_rfile import ExtendedRFile
from .nuisance_parameter_pull import NuisanceParameterPull
from .nuisance_parameter_ranking import NuisanceParameterRanking
from .nuisance_parameter_harmonizer import NuisanceParameterHarmonizer
+from .caching_nll_wrapper import CachingNLLWrapper
#from .signal_modelling import SignalModelling
import ROOT
diff --git a/quickstats/components/analysis_object.py b/quickstats/components/analysis_object.py
index e76ec9a457388c6eb963bb8c3a5faa243b758f3d..f93d420b9ee23c1a4e0dcae2bd5a02e083758f68 100644
--- a/quickstats/components/analysis_object.py
+++ b/quickstats/components/analysis_object.py
@@ -1,11 +1,13 @@
-from typing import Optional, Union, List
+from typing import Optional, Union, List, Dict
+
+import ROOT
from quickstats import AbstractObject
from quickstats.components import ExtendedModel, ExtendedMinimizer
from quickstats.components.basics import WSArgument
-from quickstats.utils.common_utils import get_class_that_defined_method
-from quickstats.interface.root import RooArgSet
-import ROOT
+from quickstats.utils.common_utils import get_class_that_defined_method, update_config_dict, combine_dict
+from quickstats.utils.roostats_utils import set_prior_pdf
+from quickstats.interface.root import RooArgSet, ModelConfig
class AnalysisObject(AbstractObject):
@@ -18,12 +20,13 @@ class AnalysisObject(AbstractObject):
fix_param:str='', profile_param:str='', ws_name:Optional[str]=None,
mc_name:Optional[str]=None, snapshot_name:Optional[Union[List[str], str]]=None,
minimizer_type:str='Minuit2', minimizer_algo:str='Migrad', precision:float=0.001,
- eps:float=1.0, retry:int=1, strategy:int=1, print_level:int=-1, timer:bool=False,
- num_cpu:int=1, offset:bool=True, optimize:int=2, eigen:bool=False, hesse:bool=False,
- improve:int=0, fix_cache:bool=True, fix_multi:bool=True, max_calls:int=-1,
- max_iters:int=-1, constrain_nuis:bool=True, batch_mode:bool=False,
+ eps:float=1.0, retry:int=1, strategy:int=1, print_level:int=-1,
+ num_cpu:int=1, offset:bool=True, optimize:int=2, constrain_nuis:bool=True,
+ batch_mode:bool=False, prior_type:Optional[str]=None, runtimedef_expr:Optional[str]=None,
int_bin_precision:float=-1., preset_param:bool=False, minimizer_offset:int=1,
- minimizer_cls=None, verbosity:Optional[Union[int, str]]="INFO"):
+ extra_minimizer_options:Optional[Union[Dict, str]]=None,
+ minimizer_cls=None, verbosity:Optional[Union[int, str]]="INFO",
+ **kwargs):
super().__init__(verbosity=verbosity)
self.model = None
self.minimizer = None
@@ -39,52 +42,74 @@ class AnalysisObject(AbstractObject):
"mc_name" : mc_name,
"data_name" : data_name,
"snapshot_name" : snapshot_name,
- "binned_likelihood" : binned_likelihood,
- "fix_cache" : fix_cache,
- "fix_multi" : fix_multi
+ "binned_likelihood" : binned_likelihood
}
self.setup_model(**model_options)
self.model._load_floating_auxiliary_variables()
self.save_snapshot(self.kInitialSnapshotName, WSArgument.MUTABLE)
- self.set_poi(poi_name)
- if preset_param:
- self.preset_parameters()
- self.setup_parameters(fix_param, profile_param, update_snapshot=True)
minimizer_options = {
+ "constrain_nuis" : constrain_nuis,
"minimizer_type" : minimizer_type,
"minimizer_algo" : minimizer_algo,
"precision" : precision,
"eps" : eps,
- "retry" : retry,
"strategy" : strategy,
- "num_cpu" : num_cpu,
- "offset" : offset,
"minimizer_offset" : minimizer_offset,
+ "offset" : offset,
"optimize" : optimize,
- "eigen" : eigen,
- "hesse" : hesse,
- "improve" : improve,
- "max_calls" : max_calls,
- "max_iters" : max_iters,
+ "retry" : retry,
+ "num_cpu" : num_cpu,
"print_level" : print_level,
- "timer" : timer,
- "constrain_nuis" : constrain_nuis,
"batch_mode" : batch_mode,
"int_bin_precision" : int_bin_precision,
+ "runtimedef_expr" : runtimedef_expr,
"verbosity" : verbosity
}
- self.setup_minimizer(**minimizer_options)
+ if kwargs:
+ minimizer_options = combine_dict(minimizer_options, kwargs)
+ minimizer_options = update_config_dict(minimizer_options, extra_minimizer_options)
+ self.setup_minimizer(discrete_nuisance=self.model._discrete_nuisance,
+ **minimizer_options)
+ self.set_poi(poi_name)
+ if preset_param:
+ self.preset_parameters()
+ self.setup_parameters(fix_param, profile_param, update_snapshot=True)
+ set_prior_pdf(self.model.workspace,
+ self.model.model_config,
+ self.poi, prior_type)
self.sanity_check()
+
def sanity_check(self):
- aux_vars = self.model.get_variables("auxiliary")
- floating_aux_vars = [v.GetName() for v in aux_vars if not v.isConstant()]
- if len(floating_aux_vars) > 0:
- self.stdout.warning("The following auxiliary variables (variables that are not "
- "part of the POIs, observables, nuisance parameters and global "
- f"observables) are floating: {','.join(floating_aux_vars)}. If this is "
- "not intended, please make sure to fix them before fitting.", "red")
-
+
+ def check_aux_vars_are_fixed():
+ aux_vars = self.model.get_variables("auxiliary")
+ ROOT.RooStats.RemoveConstantParameters(aux_vars)
+ if self.model.has_discrete_nuisance():
+ aux_vars = RooArgSet.get_set_difference(aux_vars, self.model.discrete_nuisance.multipdf_params)
+ aux_vars = RooArgSet.get_set_difference(aux_vars, ROOT.RooArgSet(self.model.discrete_nuisance.multipdf_cats))
+ if len(aux_vars) > 0:
+ aux_var_names = [v.GetName() for v in aux_vars]
+ self.stdout.warning("The following auxiliary variables (variables that are not "
+ "part of the POIs, observables, nuisance parameters, global "
+ "observables and discrete nuisances) are floating: "
+ f"{','.join(aux_var_names)}. If this is not intended, please "
+ "make sure to fix them before fitting.", "red")
+ passed = len(aux_vars) == 0
+ return passed
+
+ def check_model_config():
+ passed = ModelConfig(verbosity=self.stdout.verbosity).sanity_check(self.model.model_config)
+ return passed
+
+ for task_name, task in [('Auxiliary variable', check_aux_vars_are_fixed),
+ ('Model config', check_model_config)]:
+ passed = task()
+ if passed:
+ self.stdout.info(f"{task_name} sanity check: PASSED")
+ else:
+ self.stdout.info(f"{task_name} sanity check: FAILED")
+
def preset_parameters(self, fix_pois:bool=True, fix_globs:bool=True, float_nuis:bool=False):
if fix_pois:
ROOT.RooStats.SetAllConstant(self.model.pois, True)
@@ -111,6 +136,10 @@ class AnalysisObject(AbstractObject):
@property
def get_poi(self):
return self.model.get_poi
+
+ @property
+ def poi(self):
+ return self.minimizer._poi
# black magic
def _inherit_init(self, init_func, **kwargs):
@@ -126,23 +155,22 @@ class AnalysisObject(AbstractObject):
that_kwargs = {k:v for k,v in kwargs.items() if k in that_parameters}
this_kwargs = {k:v for k,v in kwargs.items() if k not in that_parameters}
init_func(config=this_kwargs, **that_kwargs)
-
- def set_poi(self, poi_name:Optional[Union[str, List[str]]]=None):
- pois = self.get_poi(poi_name)
+
+ def set_poi(self, pois:Optional[Union[str, List[str], "ROOT.RooRealVar", "ROOT.RooArgSet"]]=None):
+ if (pois is not None) and not isinstance(pois, (ROOT.RooRealVar, ROOT.RooArgSet)):
+ pois = self.get_poi(pois)
+ self.minimizer.set_poi(pois)
if isinstance(pois, ROOT.RooRealVar):
- poi_text = f'"{pois.GetName()}"'
- n_poi = 1
- elif isinstance(pois, ROOT.RooArgSet) and (len(pois) > 0):
- poi_text = ", ".join([f'"{poi.GetName()}"' for poi in pois])
- n_poi = len(pois)
+ poi_names = [pois.GetName()]
+ elif isinstance(pois, ROOT.RooArgSet):
+ poi_names = [poi.GetName() for poi in pois]
else:
- poi_text = None
- n_poi = 0
- if n_poi == 1:
- self.stdout.info(f'POI set to {poi_text}')
- elif n_poi > 1:
- self.stdout.info(f'POIs set to {poi_text}')
- self.poi = pois
+ poi_names = []
+ if len(poi_names) == 1:
+ self.stdout.info(f'POI set to "{poi_names[0]}"')
+ elif len(poi_names) > 1:
+ text = ", ".join([f'"{name}"' for name in poi_names])
+ self.stdout.info(f'POIs set to {text}')
def setup_model(self, **kwargs):
model = ExtendedModel(**kwargs, verbosity=self.stdout.verbosity)
@@ -196,11 +224,14 @@ class AnalysisObject(AbstractObject):
f"POI used in this study. This parameter will still be floated during "
f"unconditional likelihood fit / limit setting.", "red")
- def setup_minimizer(self, constrain_nuis:bool=True, **kwargs):
-
+ def setup_minimizer(self, constrain_nuis:bool=True,
+ discrete_nuisance:Optional["DiscreteNuisance"]=None,
+ runtimedef_expr:Optional[str]=None, **kwargs):
minimizer = self.minimizer_cls("Minimizer", self.model.pdf, self.model.data,
workspace=self.model.workspace,
- verbosity=self.stdout.verbosity)
+ runtimedef_expr=runtimedef_expr,
+ verbosity=self.stdout.verbosity,
+ discrete_nuisance=discrete_nuisance)
nll_options = {k:v for k,v in kwargs.items() if k in ExtendedMinimizer._DEFAULT_NLL_OPTION_}
@@ -216,13 +247,17 @@ class AnalysisObject(AbstractObject):
minimizer.configure_nll(**nll_options)
minimizer.configure(**minimizer_options)
self.minimizer = minimizer
+ self.model.set_minimizer(self.minimizer)
self.default_nll_options = nll_options
self.default_minimizer_options = minimizer_options
def set_data(self, data_name:str='combData'):
- data = self.model.workspace.data(data_name)
- if not data:
- raise RuntimeError(f'workspace does not contain the dataset "{data_name}"')
+ if isinstance(data_name, ROOT.RooDataSet):
+ data = data_name
+ else:
+ data = self.model.workspace.data(data_name)
+ if not data:
+ raise RuntimeError(f'workspace does not contain the dataset "{data_name}"')
self.minimizer.set_data(data)
self.model._data = data
diff --git a/quickstats/components/asymptotic_cls.py b/quickstats/components/asymptotic_cls.py
index 5150f040f10cfdc2c4c31ceace736a2b80a3ed7c..2f0ee9e87552743125d016c3a6f75721324e29d1 100644
--- a/quickstats/components/asymptotic_cls.py
+++ b/quickstats/components/asymptotic_cls.py
@@ -33,13 +33,11 @@ class AsymptoticCLs(AnalysisObject):
better_negative_bands:bool=False, do_blind:bool=True,
binned_likelihood:bool=True, fix_param:str='', profile_param:str='',
ws_name:Optional[str]=None, mc_name:Optional[str]=None,
- snapshot_name:Optional[str]=None, timer:bool=False,
+ snapshot_name:Optional[str]=None, constrain_nuis:bool=True,
minimizer_type:str='Minuit2', minimizer_algo:str='Migrad',
eps:float=1.0, retry:int=2, strategy:int=1, print_level:int=-1,
- num_cpu:int=1, offset:bool=True, optimize:int=2,
- fix_cache:bool=True, fix_multi:bool=True, max_calls:int=-1,
- max_iters:int=-1, constrain_nuis:bool=True, batch_mode:bool=False,
- int_bin_precision:float=-1., improve:int=0, minimizer_offset:int=1,
+ num_cpu:int=1, minimizer_offset:int=1, offset:bool=True,
+ optimize:int=2, batch_mode:bool=False, int_bin_precision:float=-1.,
verbosity:Optional[Union[int, str]]="INFO", **kwargs):
"""
args:
@@ -64,24 +62,18 @@ class AsymptoticCLs(AnalysisObject):
'retry': retry,
'strategy': strategy,
'num_cpu': num_cpu,
+ 'minimizer_offset': minimizer_offset,
'offset': offset,
'optimize': optimize,
- 'improve': improve,
- 'minimizer_offset': minimizer_offset,
- 'eigen': False,
'print_level': print_level,
- 'timer': timer,
- 'fix_cache': fix_cache,
- 'fix_multi': fix_multi,
- 'max_calls': max_calls,
- 'max_iters': max_iters,
'constrain_nuis': constrain_nuis,
'batch_mode': batch_mode,
'int_bin_precision': int_bin_precision,
'preset_param': True,
- 'verbosity': verbosity
+ 'verbosity': verbosity,
+ **kwargs
}
- super().__init__(**config)
+ self._inherit_init(super().__init__, **config)
self.do_blind = do_blind
self.CL = CL
self.precision = precision
@@ -98,20 +90,8 @@ class AsymptoticCLs(AnalysisObject):
self.stdout.info(f"Changed mu_guess to {mu_guess} so that it is different from mu_exp")
self.mu_guess = mu_guess
- # define asimov data
- if asimov_data_name is not None:
- asimov_data_0 = self.model.workspace.data(asimov_data_name)
- if not asimov_data_0:
- raise RuntimeError(f'failed to load dataset \"{asimov_data_name}\"')
- self.asimov_data_0 = asimov_data_0
- glob_snap_name = self.get_conditional_snapsot_name(0, "global_observable")
- nuis_snap_name = self.get_conditional_snapsot_name(0, "nuisance_parameter")
- if (not self.model.workspace.loadSnapshot(glob_snap_name)) or \
- (not self.model.workspace.loadSnapshot(nuis_snap_name)):
- raise RuntimeError("when using user-defined background-only Asimov data, "
- f"snapshots {nuis_snap_name} and {glob_snap_name} must be defined.")
- else:
- self.asimov_data_0 = None
+ # use custom bkg asimov data if specified
+ self.load_asimov_data_0(asimov_data_name)
# define simplified names
self.ws = self.model.workspace
@@ -138,22 +118,31 @@ class AsymptoticCLs(AnalysisObject):
self.stdout.info('Loaded extension module "AsymptoticCLsTool"')
except Exception as e:
print(e)
-
- def set_poi(self, poi_name:Optional[str]=None):
- if poi_name is None:
- self.poi = self.model.pois.first()
- self.stdout.info('POI name not given. Set to first poi "{}" by default.'.format(self.poi.GetName()))
- else:
- self.poi = self.model.get_poi(poi_name)
- self.stdout.info('POI set to "{}"'.format(poi_name))
+
+ def load_asimov_data_0(self, asimov_data_name:Optional[str]=None):
+ if asimov_data_name is None:
+ self.asimov_data_0 = None
+ return None
+ asimov_data_0 = self.model.workspace.data(asimov_data_name)
+ if not asimov_data_0:
+ raise RuntimeError(f'failed to load dataset "{asimov_data_name}"')
+ # check that conditional global observable and nuisance parameter snapshots are defined
+ glob_snap_name = self.get_conditional_snapsot_name(0, "global_observable")
+ nuis_snap_name = self.get_conditional_snapsot_name(0, "nuisance_parameter")
+ glob_snap = self.model.workspace.getSnapshot(glob_snap_name)
+ nuis_snap = self.model.workspace.getSnapshot(nuis_snap_name)
+ if ((not glob_snap) or (not nuis_snap)):
+ raise RuntimeError("when using user-defined background-only Asimov data, "
+ f"snapshots {nuis_snap_name} and {glob_snap_name} must be defined.")
+ self.asimov_data_0 = asimov_data_0
def set_poi_value(self, val:float):
if math.isnan(val):
raise ValueError("cannot set poi to nan")
if ( val > 0 ) and (self.poi.getMax() < val):
- self.poi.setMax(2*val)
+ self.poi.setMax(2 * val)
if ( val < 0 ) and (self.poi.getMin() > val):
- self.poi.setMin(2*val)
+ self.poi.setMin(2 * val)
self.poi.setVal(val)
def save_nll_snapshot(self, nll, mu):
@@ -186,16 +175,10 @@ class AsymptoticCLs(AnalysisObject):
self.load_snapshot(snapshot_name)
else:
raise RuntimeError(f'Failed to load snapshot for nll "{nll.GetName()}"')
- if self.stdout.verbosity <= "DEBUG":
- self.stdout.debug("Before Fit ----------------------------------------------")
- self.model.print_summary(items=['poi', 'nuisance_parameter', 'global_observable'])
- self.last_fit_status = self.minimizer.minimize(nll)
- self.global_status += self.last_fit_status
+ self.last_fit_status = self.minimizer.minimize(nll)
+ self.global_status += self.last_fit_status
self.minimizer_calls += 1
nll_val = nll.getVal()
- if self.stdout.verbosity <= "DEBUG":
- self.stdout.debug("After Fit ----------------------------------------------")
- self.model.print_summary(items=['poi', 'nuisance_parameter', 'global_observable'])
self.load_snapshot(self.nom_glob_name)
return nll_val
@@ -205,7 +188,7 @@ class AsymptoticCLs(AnalysisObject):
def get_qmu(self, nll:"ROOT.RooNLLVar", mu:float):
nll_muhat = self.nll_maps[nll].get('nll_mu_hat', None)
if nll_muhat is None:
- raise RuntimeError('Failed to get nominal nll for "{}"'.format(nll.GetName()))
+ raise RuntimeError(f'Failed to get nominal nll for "{nll.GetName()}"')
is_const = self.poi.isConstant()
self.poi.setConstant(1)
self.set_poi_value(mu)
@@ -251,10 +234,13 @@ class AsymptoticCLs(AnalysisObject):
self.stdout.info("----------------------------------", bare=True)
self.stdout.info(f"Getting limit for nll: {nll_name}", bare=True)
self.poi.setConstant(0)
+ # load mu_exp asimov
asimov_0_nll = self.summary[0]['nll']
+
if nll == asimov_0_nll:
self.set_poi_value(self.mu_exp)
self.poi.setConstant(1)
+
# case for observed limit
if (nll not in self.nll_maps) or ('nll_mu_hat' not in self.nll_maps[nll]):
nll_val = self.get_nll_val(nll)
@@ -402,9 +388,11 @@ class AsymptoticCLs(AnalysisObject):
@staticmethod
def get_approx_limits(median_limit:float, target_cls:float=0.05):
limits = {}
+ # 3.84 = value of delta chi^2 (qmu) corresponding to a coverage probability 1 − alpha = 0.95 in the large data sample limit
sigma = median_limit/math.sqrt(3.84)
limits[0] = median_limit
limits[2] = sigma*(ROOT.Math.gaussian_quantile(1 - target_cls*ROOT.Math.gaussian_cdf( 2), 1) + 2)
+
limits[1] = sigma*(ROOT.Math.gaussian_quantile(1 - target_cls*ROOT.Math.gaussian_cdf( 1), 1) + 1)
limits[-1] = sigma*(ROOT.Math.gaussian_quantile(1 - target_cls*ROOT.Math.gaussian_cdf(-1), 1) - 1)
limits[-2] = sigma*(ROOT.Math.gaussian_quantile(1 - target_cls*ROOT.Math.gaussian_cdf(-2), 1) - 2)
@@ -431,55 +419,7 @@ class AsymptoticCLs(AnalysisObject):
self.nll_maps[nll]['mu_hat'] = mu_hat
self.nll_maps[nll]['nll_mu_hat'] = nll_mu_hat
self.dataset_maps[dataset] = nll
-
- def print_failures(self):
- has_failtures = any(self.summary[sigma].get('status', 0) > 0 for sigma in self.summary)
- if has_failtures:
- self.stdout.info('-----------------------------------------------', bare=True)
- self.stdout.info('Unresolved fit failures detected', bare=True)
- for sigma in self.summary:
- label = self._SIGMA_LABEL_.get(sigma, sigma)
- self.stdout.info("{}:".format(label).ljust(10) + str(self.summary[sigma].get('status', None)), bare=True)
- self.stdout.info('-----------------------------------------------', bare=True)
- def print_summary(self):
- if (not self.limits) or (not self.approx_limits):
- self.stdout.warning("No limits evaluated")
- return None
- self.stdout.info('', bare=True)
- self.print_failures()
- if self.do_better_bands:
- self.stdout.info('Guess for bands', bare=True)
- self.stdout.info(f'+2sigma: {self.approx_limits[2]}', bare=True)
- self.stdout.info(f'+1sigma: {self.approx_limits[1]}', bare=True)
- self.stdout.info(f'-1sigma: {self.approx_limits[-1]}', bare=True)
- self.stdout.info(f'-2sigma: {self.approx_limits[-2]}', bare=True)
- if self.do_better_bands:
- self.stdout.info('\nCorrect bands', bare=True)
- self.stdout.info(f'+2sigma: {self.limits[2]}', bare=True)
- self.stdout.info(f'+1sigma: {self.limits[1]}', bare=True)
- self.stdout.info(f'-1sigma: {self.limits[-1]}', bare=True)
- self.stdout.info(f'-2sigma: {self.limits[-2]}', bare=True)
- self.stdout.info('Injected: {}'.format(self.limits['inj']), bare=True)
- self.stdout.info(f'Median: {self.limits[0]}', bare=True)
- self.stdout.info('Observed: {}'.format(self.limits['obs']), bare=True)
- self.stdout.info('', bare=True)
-
- def save(self, filename:str='limits.json', parameters:Optional[Dict]=None, summary:bool=False):
- with open(filename, "w") as file:
- if parameters is None:
- json.dump(self.limits, file, indent=2)
- else:
- json.dump({**parameters, **self.limits}, file, indent=2)
- file.truncate()
- if summary:
- data = {k:{kk:vv for kk,vv in self.summary[k].items() if kk not in ['nll', 'dataset']} \
- for k,v in self.summary.items()}
- summary_filename = os.path.splitext(filename)[0]+'_summary'+os.path.splitext(filename)[1]
- with open(summary_filename, "w") as file:
- json.dump(data, file, indent=2)
- file.truncate()
-
def evaluate_limit_band(self, n:int, med_limit:float):
init_target_cls = self.LimitTool.getTargetCLs()
self.LimitTool.setTargetCLs(2*(1 - ROOT.Math.gaussian_cdf(abs(n))))
@@ -550,11 +490,14 @@ class AsymptoticCLs(AnalysisObject):
self.save_nll_snapshot(asimov_0_nll, self.mu_exp)
self.load_conditional_snapshot(self.mu_exp, target=WSArgument.GLOBAL_OBSERVABLE)
self.load_conditional_snapshot(self.mu_exp, target=WSArgument.NUISANCE_PARAMETER)
+ if (self.minimizer.has_discrete_nuisance() and self.do_blind):
+ self.minimizer.minimize(asimov_0_nll)
nll_value = asimov_0_nll.getVal()
self.fill_mappings(asimov_0_nll, asimov_data_0, self.mu_exp, nll_value)
self.LimitTool.setTargetCLs(1 - self.CL)
-
+
+ # evaluate median expected limit
med_limit = self.get_limit(asimov_0_nll, self.mu_guess, 0)
approx_limits = self.get_approx_limits(med_limit, self.LimitTool.getTargetCLs())
limits = approx_limits.copy()
@@ -633,3 +576,51 @@ class AsymptoticCLs(AnalysisObject):
self.stdout.info(f"pmu: {pmu}", bare=True)
self.stdout.info(f"1-pb: {pb}", bare=True)
self.stdout.info(f"CLs: {CLs}", bare=True)
+
+ def print_failures(self):
+ has_failtures = any(self.summary[sigma].get('status', 0) > 0 for sigma in self.summary)
+ if has_failtures:
+ self.stdout.info('-----------------------------------------------', bare=True)
+ self.stdout.info('Unresolved fit failures detected', bare=True)
+ for sigma in self.summary:
+ label = self._SIGMA_LABEL_.get(sigma, sigma)
+ self.stdout.info("{}:".format(label).ljust(10) + str(self.summary[sigma].get('status', None)), bare=True)
+ self.stdout.info('-----------------------------------------------', bare=True)
+
+ def print_summary(self):
+ if (not self.limits) or (not self.approx_limits):
+ self.stdout.warning("No limits evaluated")
+ return None
+ self.stdout.info('', bare=True)
+ self.print_failures()
+ if self.do_better_bands:
+ self.stdout.info('Guess for bands', bare=True)
+ self.stdout.info(f'+2sigma: {self.approx_limits[2]}', bare=True)
+ self.stdout.info(f'+1sigma: {self.approx_limits[1]}', bare=True)
+ self.stdout.info(f'-1sigma: {self.approx_limits[-1]}', bare=True)
+ self.stdout.info(f'-2sigma: {self.approx_limits[-2]}', bare=True)
+ if self.do_better_bands:
+ self.stdout.info('\nCorrect bands', bare=True)
+ self.stdout.info(f'+2sigma: {self.limits[2]}', bare=True)
+ self.stdout.info(f'+1sigma: {self.limits[1]}', bare=True)
+ self.stdout.info(f'-1sigma: {self.limits[-1]}', bare=True)
+ self.stdout.info(f'-2sigma: {self.limits[-2]}', bare=True)
+ self.stdout.info('Injected: {}'.format(self.limits['inj']), bare=True)
+ self.stdout.info(f'Median: {self.limits[0]}', bare=True)
+ self.stdout.info('Observed: {}'.format(self.limits['obs']), bare=True)
+ self.stdout.info('', bare=True)
+
+ def save(self, filename:str='limits.json', parameters:Optional[Dict]=None, summary:bool=False):
+ with open(filename, "w") as file:
+ if parameters is None:
+ json.dump(self.limits, file, indent=2)
+ else:
+ json.dump({**parameters, **self.limits}, file, indent=2)
+ file.truncate()
+ if summary:
+ data = {k:{kk:vv for kk,vv in self.summary[k].items() if kk not in ['nll', 'dataset']} \
+ for k,v in self.summary.items()}
+ summary_filename = os.path.splitext(filename)[0]+'_summary'+os.path.splitext(filename)[1]
+ with open(summary_filename, "w") as file:
+ json.dump(data, file, indent=2)
+ file.truncate()
\ No newline at end of file
diff --git a/quickstats/components/caching_nll_wrapper.py b/quickstats/components/caching_nll_wrapper.py
new file mode 100644
index 0000000000000000000000000000000000000000..6be3426fba3e68a7ff02f7be1713d26e7e5b1f95
--- /dev/null
+++ b/quickstats/components/caching_nll_wrapper.py
@@ -0,0 +1,96 @@
+##################################################################################################
+# Based on https://github.com/cms-analysis/HiggsAnalysis-CombinedLimit
+# Author: Alkaid Cheng
+# Email: chi.lung.cheng@cern.ch
+##################################################################################################
+from typing import List, Optional, Union, Dict, Set, Tuple, Sequence
+
+import numpy as np
+
+import ROOT
+import cppyy
+
+from quickstats import semistaticmethod, AbstractObject
+from quickstats.utils.common_utils import combine_dict
+from quickstats.interface.cppyy import cpp_define
+
+class CachingNLLWrapper(AbstractObject):
+ """
+ Dedicated wrapper for the CachingSimNLL class from CMS
+ """
+
+ @property
+ def nll(self):
+ return self._nll
+
+ def __init__(self, nll:Optional["ROOT.RooAbsReal"]=None,
+ verbosity:Optional[Union[int, str]]="INFO"):
+ self.set_nll(nll)
+
+ @semistaticmethod
+ def is_defined_class(self):
+ return hasattr(cppyy.gbl, 'cacheutils::CachingSimNLL')
+
+ @semistaticmethod
+ def define_cast_function(self):
+ if not self.is_defined_class():
+ return None
+ expr = 'cacheutils::CachingSimNLL * castAsCachingSimNLL(RooAbsReal* nll){ return dynamic_cast<cacheutils::CachingSimNLL *>(nll);}'
+ status = cpp_define(expr, 'CachingNLLWrapperMethod')
+ return status
+
+ def set_nll(self, nll:Optional["ROOT.RooAbsReal"]=None):
+ if (nll is None) or (not self.is_defined_class()):
+ self._nll = None
+ return None
+ if not hasattr(cppyy.gbl, 'castAsCachingSimNLL'):
+ self.define_cast_function()
+ caching_nll = cppyy.gbl.castAsCachingSimNLL(nll)
+ if caching_nll:
+ self._nll = caching_nll
+ else:
+ self._nll = None
+
+ def set_zero_point(self):
+ """Offset the current NLL value to zero
+ """
+ if self.nll is None:
+ return None
+ self.stdout.debug(f'Setting zero point for the caching NLL: {self.nll.GetName()}')
+ self.nll.setZeroPoint()
+
+ def update_zero_point(self):
+ """Update offset value of the current NLL
+ """
+ if self.nll is None:
+ return None
+ self.stdout.debug(f'Updating zero point for the caching NLL: {self.nll.GetName()}')
+ self.nll.updateZeroPoint()
+
+ def clear_zero_point(self):
+ """Remove offset value of the current NLL
+ """
+ if self.nll is None:
+ return None
+ self.stdout.debug(f'Clearing zero point for the caching NLL: {self.nll.GetName()}')
+ self.nll.clearZeroPoint()
+
+ def set_hide_categories(self, value:bool=True):
+ if self.nll is None:
+ return None
+ self.nll.setHideRooCategories(value)
+
+ def set_mask_non_discrete_channels(self, value:bool=True):
+ if self.nll is None:
+ return None
+ self.nll.setMaskNonDiscreteChannels(value)
+
+ def set_hide_constants(self, value:bool=True):
+ if self.nll is None:
+ return None
+ self.nll.setHideConstants(value)
+
+ def set_mask_constraints(self, value:bool=True):
+ if self.nll is None:
+ return None
+ self.nll.setMaskConstraints(value)
\ No newline at end of file
diff --git a/quickstats/components/discrete_nuisance.py b/quickstats/components/discrete_nuisance.py
new file mode 100644
index 0000000000000000000000000000000000000000..26ea1bffa7a3f9895d311a6211e86b6eb7926532
--- /dev/null
+++ b/quickstats/components/discrete_nuisance.py
@@ -0,0 +1,232 @@
+##################################################################################################
+# Based on https://github.com/cms-analysis/HiggsAnalysis-CombinedLimit
+# Author: Alkaid Cheng
+# Email: chi.lung.cheng@cern.ch
+##################################################################################################
+from typing import List, Optional, Union, Dict, Set, Tuple, Sequence
+
+import numpy as np
+
+import ROOT
+
+import cppyy
+
+from quickstats import semistaticmethod, AbstractObject
+from quickstats.utils.common_utils import combine_dict
+from quickstats.interface.root import RooArgSet
+
+class DiscreteNuisance(AbstractObject):
+
+ _DEFAULT_CONFIG_ = {
+ "disc_set_keyword": "discreteParams"
+ }
+
+ @property
+ def multipdf_cats(self):
+ return self._multipdf_cats
+
+ @property
+ def multipdf_params(self):
+ return self._multipdf_params
+
+ @property
+ def all_params(self):
+ return self._all_params
+
+ @property
+ def multipdfs(self):
+ return self._multipdfs
+
+ @property
+ def freeze_flag(self):
+ return self._freeze_flag
+
+ def __init__(self, ws:Optional["ROOT.RooWorkspace"]=None,
+ pdf:Optional["ROOT.RooWorkspace"]=None,
+ config:Optional[Dict]=None,
+ verbosity:Optional[Union[int, str]]="INFO"):
+ super().__init__(verbosity=verbosity)
+ self.config = combine_dict(self._DEFAULT_CONFIG_, config)
+ self.initialize(ws, pdf)
+
+ @semistaticmethod
+ def extract_discrete_variables(self, ws:"ROOT.RooWorkspace", pdf:"ROOT.RooAbsPdf",
+ keyword:Optional[str]=None):
+ all_cats = ws.allCats().Clone()
+ if pdf.InheritsFrom("RooSimultaneous"):
+ index_cat = pdf.indexCat()
+ if index_cat in all_cats:
+ all_cats.remove(index_cat)
+ if not all_cats:
+ return ROOT.RooArgList(), ROOT.RooArgList(), ROOT.RooArgSet()
+ if (keyword is not None) and ws.genobj(keyword):
+ disc_obj = ws.genobj(keyword)
+ # sanity check
+ if not isinstance(disc_obj, ROOT.RooArgSet):
+ raise RuntimeError(f'discrete parameter container "{keyword}" is not an instance of RooArgSet')
+ disc_cats = RooArgSet.select_by_class(disc_obj, 'RooCategory')
+ if len(disc_cats) != len(disc_obj):
+ raise RuntimeError(f'discrete parameter set "{keyword}" contain instance(s) other than RooCategory')
+ else:
+ disc_cats = all_cats
+ all_pdfs = ws.allPdfs()
+ disc_pdfs = RooArgSet.select_by_class(all_pdfs, 'RooMultiPdf')
+ if not disc_pdfs:
+ return ROOT.RooArgList(), ROOT.RooArgList(), ROOT.RooArgSet()
+ dpd_disc_cats = RooArgSet.select_dependent_parameters(disc_cats, pdf)
+ dpd_disc_pdfs = RooArgSet.select_dependent_parameters(disc_pdfs, pdf)
+ if len(dpd_disc_cats) != len(dpd_disc_pdfs):
+ raise RuntimeError('mismatch between number of discrete categories and number of multi pdfs')
+ valid_disc_cats = ROOT.RooArgList()
+ valid_disc_pdfs = ROOT.RooArgList()
+ valid_disc_params = ROOT.RooArgSet()
+ for cat in dpd_disc_cats:
+ clients = ROOT.RooArgSet(*cat.clients())
+ candidate_pdfs = clients .selectCommon(dpd_disc_pdfs)
+ if len(candidate_pdfs) == 1:
+ valid_disc_cats.add(cat)
+ valid_disc_pdfs.add(candidate_pdfs)
+ elif len(candidate_pdfs) == 0:
+ raise RuntimeError(f'failed to find multi pdf associated with the category "{cat.GetName()}"')
+ else:
+ raise RuntimeError(f'failed to more than one multi pdfs associated with the category "{cat.GetName()}": '
+ f'{", ".join([pdf.GetName() for pdf in candidate_pdfs])}')
+ disc_pdf = candidate_pdfs.first()
+ disc_params = disc_pdf.getParameters(0)
+ disc_params = RooArgSet.select_by_class(disc_params, 'RooRealVar')
+ ROOT.RooStats.RemoveConstantParameters(disc_params)
+ valid_disc_params.add(disc_params)
+ return valid_disc_cats, valid_disc_pdfs, valid_disc_params
+
+ def initialize(self, ws:"ROOT.RooWorkspace", pdf:"ROOT.RooAbsPdf"):
+ if (ws is None) and (pdf is None):
+ self._multipdf_cats = ROOT.RooArgList()
+ self._multipdfs = ROOT.RooArgList()
+ self._multipdf_params = ROOT.RooArgSet()
+ self._all_params = ROOT.RooArgSet()
+ self._freeze_flag = False
+ return
+ disc_set_keyword = self.config['disc_set_keyword']
+ disc_cats, disc_pdfs, disc_params = self.extract_discrete_variables(ws, pdf, disc_set_keyword)
+ if disc_cats:
+ n_disc_cats = len(disc_cats)
+ self.stdout.info(f'Found {n_disc_cats} discrete nuisances.')
+ else:
+ self.stdout.info('No discrete nuisances found.')
+
+ self._multipdf_cats = disc_cats
+ self._multipdfs = disc_pdfs
+ self._multipdf_params = disc_params
+ self._all_params = ws.allVars()
+ self._freeze_flag = True
+
+ self.print_active_pdf_summary()
+
+ def has_discrete_nuisance(self):
+ return len(self.multipdfs) > 0
+
+ def print_active_pdf_summary(self):
+ if not self.has_discrete_nuisance():
+ return
+ self.stdout.info('Summary of multipdfs and their corresponding active pdfs:')
+ for multipdf, multipdf_cat in zip(self.multipdfs, self.multipdf_cats):
+ multipdf_name = multipdf.GetName()
+ current_pdf = multipdf.getCurrentPdf()
+ current_pdf_name = current_pdf.GetName()
+ current_index = multipdf_cat.getCurrentIndex()
+ self.stdout.info(f' {multipdf_name} -> {current_pdf_name} (index = {current_index})', bare=True)
+
+ def set_freeze_flag(self, flag:bool=True):
+ self._freeze_flag = flag
+
+ def freeze_discrete_params(self, freeze:bool=True):
+ if (not self.has_discrete_nuisance()) or (not self.freeze_flag):
+ return None
+ multipdfs = self.multipdfs
+ multipdf_params = ROOT.RooArgSet(self.multipdf_params)
+ # For each multiPdf, get the active pdf and remove its parameters
+ # from this list of params and then freeze the remaining ones
+ for multipdf in multipdfs:
+ current_pdf = multipdf.getCurrentPdf()
+ pdf_params = current_pdf.getParameters(0)
+ ROOT.RooStats.RemoveConstantParameters(pdf_params)
+ multipdf_params.remove(pdf_params)
+ num_params = len(multipdf_params)
+ if freeze:
+ self.stdout.debug(f'Freezing {num_params} disassociated multipdf parameters.')
+ else:
+ self.stdout.debug(f'Unfreezing {num_params} disassociated multipdf parameters.')
+ if self.stdout.verbosity <= "DEBUG":
+ multipdf_params.Print("V")
+ RooArgSet.set_constant_state(multipdf_params, freeze)
+
+ def get_default_pdf_indices(self):
+ return np.zeros(len(self.multipdf_cats))
+
+ def get_current_pdf_indices(self):
+ return np.array([cat.getCurrentIndex() for cat in self.multipdf_cats])
+
+ def get_pdf_sizes(self):
+ return np.array([len(cat) for cat in self.multipdf_cats])
+
+ def get_n_orthogonal_combination(self):
+ pdf_sizes = self.get_pdf_sizes()
+ return np.sum(pdf_sizes) - len(pdf_sizes) + 1
+
+ def get_n_combination(self, contributing_indices:Optional[List[np.ndarray]]=None):
+ if contributing_indices is None:
+ return np.prod(self.get_pdf_sizes())
+ return np.prod([np.sum(indices, dtype=int) for indices in contributing_indices])
+
+ def get_orthogonal_combinations(self):
+ pdf_sizes = self.get_pdf_sizes()
+ n_pdf = len(pdf_sizes)
+ combinations = np.zeros((np.sum(pdf_sizes) - n_pdf + 1, n_pdf), dtype=np.int32)
+ start_idx = 1
+ for i, size in enumerate(pdf_sizes):
+ combinations[start_idx:start_idx + size - 1, i] = np.arange(1, size, dtype=np.int32)
+ start_idx += (size - 1)
+ return combinations
+
+ def get_total_combinations(self, contributing_indices:Optional[List[np.ndarray]]=None):
+ pdf_sizes = self.get_pdf_sizes()
+ if contributing_indices is None:
+ grid_points = [np.arange(size, dtype=np.int32) for size in pdf_sizes]
+ else:
+ grid_points = [np.arange(size, dtype=np.int32)[np.array(contributing_indices[i], dtype=bool)] for i, size in enumerate(pdf_sizes)]
+ combinations = np.array(np.meshgrid(*grid_points)).T.reshape(-1, len(pdf_sizes))
+ return combinations
+
+ def reorder_combinations(self, combinations:np.ndarray, reference_indices:np.ndarray):
+ pdf_sizes = self.get_pdf_sizes()
+ return (combinations + reference_indices) % pdf_sizes
+
+ def create_contributing_indices(self):
+ return [np.ones(size) for size in self.get_pdf_sizes()]
+
+ def filter_combinations(self, combinations, contributing_indices:Optional[List[np.ndarray]]=None):
+ if contributing_indices is None:
+ return combinations
+ if np.all(contributing_indices == 1):
+ return combinations
+ pdf_sizes = self.get_pdf_sizes()
+ n_pdf = len(pdf_sizes)
+ max_size = np.max(pdf_sizes)
+ regular_array = np.zeros((len(pdf_sizes), max_size))
+ for i in range(n_pdf):
+ regular_array[i, 0 : pdf_sizes[i]] = contributing_indices[i]
+ valid_idx = np.where(np.choose(combinations, regular_array.T).sum(axis=1) == n_pdf)
+ return combinations[valid_idx]
+
+ def float_all_cats(self):
+ RooArgSet.set_constant_state(self.multipdf_cats, False)
+
+ def fix_non_target_cats(self, target_index:int):
+ if target_index < 0:
+ RooArgSet.set_constant_state(self.multipdf_cats, False)
+ else:
+ RooArgSet.set_constant_state(self.multipdf_cats, True)
+ self.multipdf_cats.at(target_index).setConstant(False)
+
+ def set_category_indices(self, indices:np.ndarray):
+ return RooArgSet.set_category_indices(self.multipdf_cats, indices)
\ No newline at end of file
diff --git a/quickstats/components/extended_minimizer.py b/quickstats/components/extended_minimizer.py
index 1ffbad2b81b83688f691f7eac5e262e74eed1e62..066b65277b6ad9df6b1a89489ac8ddadd62a2603 100644
--- a/quickstats/components/extended_minimizer.py
+++ b/quickstats/components/extended_minimizer.py
@@ -9,8 +9,16 @@ import math
import numpy as np
import ROOT
+import cppyy
+
import quickstats
-from quickstats import AbstractObject
+from quickstats import AbstractObject, Timer
+from quickstats.utils.string_utils import split_str
+from quickstats.utils.common_utils import combine_dict
+from quickstats.interface.root import RooArgSet
+from quickstats.interface.cppyy.vectorize import as_vector
+from .discrete_nuisance import DiscreteNuisance
+from .caching_nll_wrapper import CachingNLLWrapper
class ExtendedMinimizer(AbstractObject):
_SCAN_EXCLUDE_CONFIG_ = ['scan', 'minos', 'save']
@@ -41,7 +49,24 @@ class ExtendedMinimizer(AbstractObject):
'n_sigma': 1,
'precision': 0.001,
'strategy': 1,
- 'print_level': -1
+ 'print_level': -1,
+ 'error_level': -1,
+ # extra configs
+ 'check_boundary': 1,
+ 'prefit_hesse': 0,
+ 'postfit_hesse': 0,
+ 'minuit2_storage_level': 0,
+ # cms dedicated configs
+ 'set_zero_point': 1, # change the reference point of the NLL to be zero during minimization
+ # discrete nuisance configs
+ 'discrete_min_tol': 0.001,
+ 'do_discrete_iteration': 1,
+ 'freeze_disassociated_params': 1,
+ 'do_short_combination': 1,
+ 'max_short_combination_iteration': 15,
+ 'multimin_hide_constants': 1,
+ 'multimin_mask_constraints': 1,
+ 'multimin_mask_channels': 2
}
_DEFAULT_NLL_OPTION_ = {
@@ -50,13 +75,35 @@ class ExtendedMinimizer(AbstractObject):
'batch_mode': False,
'int_bin_precision': -1
}
+
+ _DEFAULT_RUNTIMEDEF_ = {
+ "OPTIMIZE_BOUNDS": 1,
+ "ADDNLL_RECURSIVE": 1,
+ "ADDNLL_GAUSSNLL": 1,
+ "ADDNLL_HISTNLL": 1,
+ "ADDNLL_CBNLL": 1,
+ "TMCSO_AdaptivePseudoAsimov": 1,
+ "ADDNLL_ROOREALSUM_FACTOR": 1,
+ "ADDNLL_ROOREALSUM_NONORM": 1,
+ "ADDNLL_ROOREALSUM_BASICINT": 1,
+ "ADDNLL_ROOREALSUM_KEEPZEROS": 1,
+ "ADDNLL_PRODNLL": 1,
+ "ADDNLL_HFNLL": 1,
+ "ADDNLL_HISTFUNCNLL": 1,
+ "ADDNLL_ROOREALSUM_CHEAPPROD": 1
+ }
@property
def config(self):
return self._config
- def __init__(self, minimizer_name:str, pdf:"ROOT.RooAbsPdf", data:"ROOT.RooAbsData",
+ def __init__(self, minimizer_name:str="Minimizer",
+ pdf:Optional["ROOT.RooAbsPdf"]=None,
+ data:Optional["ROOT.RooAbsData"]=None,
+ poi:Optional["ROOT.RooArgSet"]=None,
minimizer_options:Optional[Dict]=None,
+ runtimedef_expr:Optional[str]=None,
+ discrete_nuisance:Optional[DiscreteNuisance]=None,
verbosity:Optional[Union[int, str]]="INFO",
**kwargs):
super().__init__(verbosity=verbosity)
@@ -91,6 +138,16 @@ class ExtendedMinimizer(AbstractObject):
self.nll_commands = {}
self.configure_default_minimizer_options()
+ self.set_pdf(pdf)
+ self.set_data(data)
+ self.set_poi(poi)
+
+ # CMS specific implementations
+ self.set_discrete_nuisance(discrete_nuisance)
+ self.caching_nll = CachingNLLWrapper(verbosity=self.stdout.verbosity)
+ self.set_pois_for_auto_bounds()
+ self.set_pois_for_max_bounds()
+ self.set_runtimedef(runtimedef_expr)
self.stdout.info(f'Created ExtendedMinimizer("{self.name}") instance')
@@ -103,6 +160,7 @@ class ExtendedMinimizer(AbstractObject):
assert isinstance(val, ROOT.RooAbsData)
self._data = val
self.nll = None
+ self.minimizer = None
@property
def pdf(self):
@@ -111,12 +169,30 @@ class ExtendedMinimizer(AbstractObject):
@pdf.setter
def pdf(self, val):
assert isinstance(val, ROOT.RooAbsPdf)
- self._pdf = val
+ self._pdf = val
+ self.nll = None
+ self.minimizer = None
+
+ @property
+ def discrete_nuisance(self):
+ return self._discrete_nuisance
+
+ @property
+ def poi(self):
+ return self._poi
+
+ @property
+ def auto_bounds_pois(self):
+ return self._auto_bounds_pois
+
+ @property
+ def max_bounds_pois(self):
+ return self._max_bounds_pois
@staticmethod
def _configure_default_minimizer_options(minimizer_type='Minuit2', minimizer_algo='Migrad',
strategy=0, print_level=-1,
- debug_mode=False):
+ minuit2_storage_level=0, debug_mode=False):
ROOT.Math.MinimizerOptions.SetDefaultMinimizer(minimizer_type, minimizer_algo)
ROOT.Math.MinimizerOptions.SetDefaultStrategy(strategy)
ROOT.Math.MinimizerOptions.SetDefaultPrintLevel(print_level)
@@ -127,17 +203,68 @@ class ExtendedMinimizer(AbstractObject):
ROOT.RooMsgService.instance().setGlobalKillBelow(ROOT.RooFit.FATAL)
if ROOT.Math.MinimizerOptions.DefaultPrintLevel() < 0:
ROOT.RooMsgService.instance().setGlobalKillBelow(ROOT.RooFit.FATAL)
+ if minimizer_type == 'Minuit2':
+ options = ROOT.Math.MinimizerOptions.Default('Minuit2')
+ options.SetValue('StorageLevel', minuit2_storage_level)
return None
def configure_default_minimizer_options(self):
self._configure_default_minimizer_options(self.config['minimizer_type'],
self.config['minimizer_algo'],
self.config['strategy'],
- self.config['print_level'])
+ self.config['print_level'],
+ self.config['minuit2_storage_level'])
return None
+ def set_discrete_nuisance(self, discrete_nuisance:Optional[DiscreteNuisance]=None):
+ if discrete_nuisance is None:
+ self._discrete_nuisance = DiscreteNuisance()
+ else:
+ self._discrete_nuisance = discrete_nuisance
+
+ def has_discrete_nuisance(self):
+ return self.discrete_nuisance.has_discrete_nuisance()
+
+ def set_runtimedef(self, expr:Optional[Union[str, Dict]]=None, ignore_if_not_applicable:bool=True):
+ if not hasattr(cppyy.gbl, 'runtimedef'):
+ if not ignore_if_not_applicable:
+ raise RuntimeError('can not set minimizer runtime defines: namespace "runtimedef" not defined')
+ return None
+ runtimedef_map = combine_dict(self._DEFAULT_RUNTIMEDEF_)
+ if expr is None:
+ pass
+ elif isinstance(expr, str):
+ tokens = split_str(expr, ',', remove_empty=True)
+ for token in tokens:
+ runtimedef = split_str(token, '=')
+ if (len(runtimedef) != 2) or (not runtimedef[1].isdigit()):
+ raise ValueError(f'invalid token in runtimedef expression: "{token}"')
+ name, value = runtimedef[0], int(runtimedef[1])
+ runtimedef_map[name] = value
+ elif isinstance(expr, dict):
+ runtimedef_map = combine_dict(runtimedef_map, expr)
+ else:
+ raise ValueError(f'invalid runtimedef expression: "{expr}"')
+ for name, value in runtimedef_map.items():
+ cppyy.gbl.runtimedef.set(name, value)
+ self.stdout.info(f'Set runtimedef "{name}" = {value}')
+
+ def set_pdf(self, pdf:"ROOT.RooAbsPdf"):
+ self.pdf = pdf
+
def set_data(self, data:"ROOT.RooAbsData"):
self.data = data
+
+ def set_poi(self, pois:Optional[Union["ROOT.RooRealVar", "ROOT.RooArgSet"]]=None):
+ if pois is None:
+ pois = ROOT.RooArgSet()
+ self._poi = pois
+
+ def set_pois_for_auto_bounds(self, pois:Optional["ROOT.RooArgSet"]=None):
+ self._auto_bounds_pois = pois
+
+ def set_pois_for_max_bounds(self, pois:Optional["ROOT.RooArgSet"]=None):
+ self._max_bounds_pois = pois
def _bug_fix_create_nll(self, dataset:"ROOT.RooDataSet", nll_commands:List["ROOT.RooCmdArg"]):
# fix range bug
@@ -164,6 +291,8 @@ class ExtendedMinimizer(AbstractObject):
def _create_nll(self, nll_commands:List["ROOT.RooCmdArg"]=None,
dataset:Optional["ROOT.RooDataSet"]=None):
+ if self.pdf is None:
+ raise RuntimeError('pdf not initialized')
if nll_commands is None:
nll_commands = list(self.nll_commands.values())
#self.nll = self.pdf.createNLL(self.data, nll_command_list)
@@ -172,6 +301,8 @@ class ExtendedMinimizer(AbstractObject):
range_name = command.getString(0)
self.stdout.info(f"Using the range \"{range_name}\" for NLL calculation")
if dataset is None:
+ if not self.data:
+ raise RuntimeError('dataset not initialized')
dataset = self.data
if quickstats.root_version >= (6, 26, 0):
self._bug_fix_create_nll(dataset, nll_commands)
@@ -219,11 +350,16 @@ class ExtendedMinimizer(AbstractObject):
if not self.nll:
raise RuntimeError('NLL not initialized')
+ self.caching_nll.set_nll(self.nll)
+ self.caching_nll.set_hide_categories(True)
+
if (self.minimizer is None) or (not self.config['reuse_minimizer']):
self.configure_default_minimizer_options()
self.minimizer = ROOT.RooMinimizer(self.nll)
self.minimizer.setOffsetting(self.config['minimizer_offset'])
self.minimizer.setPrintLevel(self.config['print_level'])
+ if self.config['error_level'] >= 0:
+ self.minimizer.setErrorLevel(self.config['error_level'])
self.minimizer.optimizeConst(self.config['optimize'])
self.minimizer.setMinimizerType(self.config['minimizer_type'])
self.minimizer.setEvalErrorWall(self.config['do_ee_wall'])
@@ -241,6 +377,8 @@ class ExtendedMinimizer(AbstractObject):
if self.config['max_iters'] != -1:
self.minimizer.setMaxIterations(self.config['max_iters'])
+
+ self.caching_nll.set_hide_categories(False)
def construct_nll_commands(self, constrain=None, global_observables=None,
conditional_observables=None, **kwargs):
@@ -267,11 +405,6 @@ class ExtendedMinimizer(AbstractObject):
command_dict = {}
for command in commands:
command_dict[command.GetName()] = command
- # temporary fix for ROOT version below 6.22/06
- #nll_command_list = ROOT.RooLinkedList()
- #for command in commands:
- # nll_command_list.Add(command)
- #return nll_command_list
return command_dict
def set_nll_commands(self, nll_commands:List["ROOT.RooCmdArg"]):
@@ -317,128 +450,177 @@ class ExtendedMinimizer(AbstractObject):
self.fit_options = kwargs
self.scan_options = {k:v for k,v in kwargs.items() if k not in self._SCAN_EXCLUDE_CONFIG_}
-
- def minimize(self, nll=None, **kwargs):
+
+ def expand_poi_bounds(self, threshold:float=0.1):
+ if (not self.auto_bounds_pois) and (not self.max_bounds_pois):
+ return False
+ expanded_bounds = False
+ for bound_type, pois in [('max', self.max_bounds_pois),
+ ('both', self.auto_bounds_pois)]:
+ if not pois:
+ continue
+ orig_pois_at_min = ROOT.RooArgSet()
+ new_pois_at_min = ROOT.RooArgSet()
+ orig_pois_at_max = ROOT.RooArgSet()
+ new_pois_at_max = ROOT.RooArgSet()
+ if bound_type == "max":
+ RooArgSet.expand_parameters_range(pois, threshold, True, False,
+ orig_pois_at_min,
+ new_pois_at_min,
+ orig_pois_at_max,
+ new_pois_at_max)
+ elif bound_type == 'both':
+ RooArgSet.expand_parameters_range(pois, threshold, True, True,
+ orig_pois_at_min,
+ new_pois_at_min,
+ orig_pois_at_max,
+ new_pois_at_max)
+ else:
+ raise ValueError(f'unknown bound type: {bound_type}')
+ threshold_pc = threshold * 100
+ for orig_poi, new_poi in zip(orig_pois_at_min, new_pois_at_min):
+ expanded_bounds = True
+ self.stdout.info(f'Parameter {orig_poi.GetName()} has value {orig_poi.getVal()} which is within '
+ f'{threshold_pc}% from the low boundary {orig_poi.getMin()}. Will enlarge range '
+ f'to [{new_poi.getMin()}, {new_poi.getMax()}].')
+ for orig_poi, new_poi in zip(orig_pois_at_max, new_pois_at_max):
+ expanded_bounds = True
+ self.stdout.info(f'Parameter {orig_poi.GetName()} has value {orig_poi.getVal()} which is within '
+ f'{threshold_pc}% from the max boundary {orig_poi.getMax()}. Will enlarge range '
+ f'to [{new_poi.getMin()}, {new_poi.getMax()}].')
+ return expanded_bounds
+
+ def check_param_boundary(self):
+ if not self.nll:
+ self.stdout.warning('Failed to check boundary values: nll not set')
+ return None
+ nll_params = self.nll.getParameters(self.data.get())
+ nll_params.remove(self.poi)
+ ROOT.RooStats.RemoveConstantParameters(nll_params)
+ nll_params = RooArgSet.select_by_class(nll_params, 'RooRealVar')
+ boundary_params = RooArgSet.get_boundary_parameters(nll_params)
+ if boundary_params:
+ self.stdout.warning('Found parameters near boundary (within 1 sigma) after fit.')
+ for param in boundary_params:
+ self.stdout.warning(f' Parameter = {param.GetName()}, Value = {param.getVal()}, '
+ f'RangeLo = {param.getMin()}, RangeHi = {param.getMax()}, '
+ f'ErrorLo = {param.getErrorLo()}, ErrorHi = {param.getErrorHi()}',
+ bare=True)
+
+ def save_fit_result(self, named:bool=False):
+ if not self.minimizer:
+ self.stdout.warning('Failed to save fit result: minimizer not set')
+ return None
+ if named:
+ data_name = self.data.GetName()
+ save_name = f"fitresult_{self.name}_{data_name}"
+ save_title = f"Result of fit of p.d.f. {self.name} to dataset {data_name}"
+ self.stdout.info(f'ExtendedMinimizer::minimize("{self.name}") saving results as {save_name}')
+ self.fit_result = self.minimizer.save(save_name, save_title)
+ else:
+ self.fit_result = self.minimizer.save()
+
+ def minimize(self, nll=None, cascade:bool=True, **kwargs):
self.configure(**kwargs)
if nll is None:
self.create_nll()
else:
self.nll = nll
-
- if self.cond_set.size() > 0:
- attached_set = self.nll.getVariables()
- for cond in self.cond_set:
- buffer = attached_set.find(cond.GetName())
- if buffer:
- buffer.setVal(cond.getVal())
- buffer.setConstant(1)
+ self.min_nll = None
+ self.discrete_nuisance.set_freeze_flag(self.config['freeze_disassociated_params'])
+
+ if self.cond_set:
+ nll_variables = self.nll.getVariables()
+ selected_variables = nll_variables.selectCommon(self.cond_set)
+ selected_variables.assignValueOnly(self.cond_set)
+ RooArgSet.set_constant_state(selected_variables, True)
self.create_minimizer()
+
status = 0
- attached_set = self.nll.getVariables()
- perform_minimization = any(not attached.isConstant() for attached in attached_set)
-
+ nll_variables = self.nll.getVariables()
+ perform_minimization = RooArgSet.select_by_constant_state(nll_variables, False)
if not perform_minimization:
- self.stdout.info('ExtendedMinimizer::minimize("{}") no floating parameters found'
- '-- skipping minimization'.format(self.name))
+ self.stdout.info('ExtendedMinimizer::minimize("{}") no floating parameters found '
+ '-- skipping minimization'.format(self.name))
+ self.min_nll = self.nll.getVal()
else:
- status = self.robust_minimize()
+ self.discrete_nuisance.freeze_discrete_params(True)
- # Evaluate errors with improve
- if self.config['improve']:
- self.minimizer.improve()
-
- # Evaluate errors with Hesse
- if self.config['hesse']:
- self.minimizer.hesse()
-
- # Obtain Hessian matrix either from patched Minuit or after inversion
- # TMatrixDSym G = Minuit2::MnHesse::lastHessian();
- last_fit = self.minimizer.lastMinuitFit()
- if last_fit:
- self.Hessian_matrix = last_fit.covarianceMatrix().Invert()
- else:
- self.Hessian_matrix = None
-
- # Eigenvalue and eigenvector analysis
- if self.config['eigen']:
- self.eigen_analysis()
-
- # Evaluate errors with Minos
- if self.config['minos']:
- if self.minos_set.size() > 0:
- self.minimizer.minos(self.minos_set)
+ # actual minimization done here
+ if (self.has_discrete_nuisance() and self.config['do_discrete_iteration']):
+ status = self.discrete_minimize(cascade=cascade)
else:
- self.minimizer.minos()
+ status = self.robust_minimize(cascade=cascade)
+
+ # Evaluate errors with improve
+ if self.config['improve']:
+ self.minimizer.improve()
+
+ # Evaluate errors with Hesse
+ if self.config['hesse']:
+ self.minimizer.hesse()
+
+ # Eigenvalue and eigenvector analysis
+ if self.config['eigen']:
+ self.eigen_analysis()
+
+ # Evaluate errors with Minos
+ if self.config['minos']:
+ if self.minos_set.size() > 0:
+ self.minimizer.minos(self.minos_set)
+ else:
+ self.minimizer.minos()
+
+ self.discrete_nuisance.freeze_discrete_params(False)
+
+ # post-processing
+ if self.config['check_boundary']:
+ self.check_param_boundary()
- self.min_nll = self.nll.getVal()
+ self.min_nll = self.nll.getVal()
- if self.config['scan']:
- self.find_sigma()
+ if self.config['scan']:
+ self.find_sigma()
- if self.config['save']:
- data_name = self.data.GetName()
- save_name = "fitresult_{}_{}".format(self.name, data_name)
- save_title = "Result of fit of p.d.f. {} to dataset {}".format(self.name, data_name)
- self.stdout.info('ExtendedMinimizer::minimize("{}") saving results as {}'.format(self.name, save_name))
- self.fit_result = self.minimizer.save(save_name, save_title)
- # if the 'scan' option is used, minimizer will get deleted before this line
- elif self.minimizer is not None:
- self.fit_result = self.minimizer.save()
-
- if self.cond_set.size() > 0:
- attached_set = self.nll.getVariables()
- for cond in self.cond_set:
- buffer = attached_set.find(cond.GetName())
- if buffer:
- buffer.setVal(cond.getVal())
- buffer.setConstant(cond.isConstant())
+ # if the 'scan' option is used, minimizer will get deleted before this line
+ if self.minimizer is not None:
+ self.save_fit_result(named=self.config['save'])
+
+ if self.cond_set:
+ nll_variables = self.nll.getVariables()
+ selected_variables = nll_variables.selectCommon(self.cond_set)
+ selected_variables.assign(self.cond_set)
+ # dispose of minimizer
if not self.config['reuse_minimizer']:
self.minimizer = None
self.status = status
return status
-
- def eigen_analysis(self):
- if not isinstance(self.Hessian_matrix, ROOT.TMatrixDSym):
- raise ValueError('invalid Hessian matrix')
- n = self.Hessian_matrix.GetNrows()
-
- # construct reduced Hessian matrix
- Gred = ROOT.TMatrixDSym(n)
- for i in range(n):
- for j in range(n):
- norm = math.sqrt(self.Hessian_matrix(i, i)*self.Hessian_matrix(j, j))
- Gred[i][j] = self.Hessian_matrix(i, j)/norm
-
- # perform eigenvalue analysis using ROOT standard tools
- Geigen = ROOT.TMatrixDSymEigen(Gred)
-
- self.eigen_values = Geigen.GetEigenValues()
- self.eigen_vectors = Geigen.GetEigenVectors()
-
- # simple printing of eigenvalues and eigenvectors
- self.eigen_values.Print()
- self.eigen_vectors.Print()
-
- def robust_minimize(self):
+ def robust_minimize(self, cascade:bool=True):
+ self.stdout.debug('Begin robust minimization.')
strategy = self.config['strategy']
retry = self.config['retry']
minimizer_type = self.config['minimizer_type']
minimizer_algo = self.config['minimizer_algo']
- status = self.minimizer.minimize(minimizer_type, minimizer_algo)
- while ((status != 0) and (status!=1) and (retry>0)):
- if (strategy < 2):
+ status = self.single_minimize(minimizer_type, minimizer_algo)
+ # repeat if fit failed or poi(s) at boundary
+ expanded_pois = self.expand_poi_bounds()
+ def require_cascade(fit_status:int):
+ return cascade and (fit_status not in [0, 1])
+ while ((require_cascade(status) or expanded_pois) and (retry > 0)):
+ if (not expanded_pois) and (strategy < 2):
strategy += 1
retry -= 1
self.stdout.error(f'ExtendedMinimizer::robust_minimize("{self.name}") fit failed with status {status}. '
f'Retrying with strategy {strategy}')
self.minimizer.setStrategy(strategy)
- status = self.minimizer.minimize(minimizer_type, minimizer_algo)
+ status = self.single_minimize(minimizer_type, minimizer_algo)
+ expanded_pois = self.expand_poi_bounds()
if status not in [0, 1]:
self.stdout.error(f'ExtendedMinimizer::robust_minimize("{self.name}") fit failed with status {status}')
@@ -446,7 +628,237 @@ class ExtendedMinimizer(AbstractObject):
self.minimizer.setStrategy(self.config['strategy'])
return status
+
+ def single_minimize(self, minimizer_type:str, minimizer_algo:str):
+ self.stdout.debug('Begin single minimization.')
+ if self.minimizer is None:
+ self.create_minimizer()
+ self.discrete_nuisance.freeze_discrete_params(True)
+ if self.config['set_zero_point']:
+ self.caching_nll.set_nll(self.nll)
+ else:
+ self.caching_nll.set_nll()
+ self.caching_nll.set_zero_point()
+ if self.config['prefit_hesse']:
+ self.minimizer.hesse()
+ self.caching_nll.update_zero_point()
+ #nll_params = self.nll.getParameters(0)
+ #RooArgSet.save_data_as_txt(nll_params, 'minimizer_prefit_result.txt', 8)
+ status = self.minimizer.minimize(minimizer_type, minimizer_algo)
+ #RooArgSet.save_data_as_txt(nll_params, 'minimizer_postfit_result.txt', 8)
+ if self.config['postfit_hesse']:
+ self.caching_nll.update_zero_point()
+ self.minimizer.hesse()
+ self.caching_nll.clear_zero_point()
+ self.discrete_nuisance.freeze_discrete_params(False)
+ return status
+
+ def discrete_minimize(self, cascade:bool=True):
+ """Minimization involving discrete nuisances
+ """
+ status = 0
+
+ do_short_combination = self.config['do_short_combination']
+ if do_short_combination:
+ self.stdout.info('Begin discrete minimization with short combinations.')
+ with Timer() as t:
+ status = self.robust_minimize(cascade=cascade)
+ self.min_nll = self.nll.getVal()
+ prev_nll = self.min_nll
+ max_iteration = self.config['max_short_combination_iteration']
+ min_tol = self.config['discrete_min_tol']
+ for i in range(max_iteration):
+ self.stdout.info(f'Current iteration: {i + 1}')
+ status = self.iterative_minimize(cascade=cascade)
+ delta_nll = abs(prev_nll - self.min_nll)
+ self.stdout.info(f'Previous NLL = {prev_nll}, Minimum NLL = {self.min_nll}, Delta NLL = {delta_nll}')
+ if delta_nll < min_tol:
+ self.stdout.info(f'Delta NLL is within the required tolerence of {min_tol}.')
+ break
+ prev_nll = self.min_nll
+ num_iteration = i + 1
+ self.stdout.info(f'Discrete minimization finished in {num_iteration} iterations. Total time taken: {t.interval:.3f} s')
+ else:
+ self.stdout.info('Begin discrete minimization with full combinations.')
+ with Timer() as t:
+ clean_snapshot = ROOT.RooArgSet()
+ nll_params = self.nll.getParameters(0)
+ nll_params.remove(self.discrete_nuisance.multipdf_cats)
+ nll_params.snapshot(clean_snapshot)
+ min_nll = 10 + nll.getVal()
+ self.multiple_minimize(mode=0, clean_snapshot=clean_snapshot, cascade=cascade)
+ if len(self.discrete_nuisance.multipdfs) > 1:
+ self.multiple_minimize(mode=1, clean_snapshot=clean_snapshot, cascade=cascade)
+ self.multiple_minimize(mode=2, clean_snapshot=clean_snapshot, cascade=cascade)
+ if self.discrete_nuisance.freeze_flag:
+ self.discrete_nuisance.freeze_discrete_params(True)
+ status = self.robust_minimize(cascade=cascade)
+ self.discrete_nuisance.freeze_discrete_params(False)
+ self.stdout.info(f'Discrete minimization finished. Total time taken: {t.interval:.3f} s')
+ final_combination = self.discrete_nuisance.get_current_pdf_indices()
+ self.stdout.info(f'Final index combination: {list(final_combination)}')
+ return status
+
+ # taken from https://github.com/cms-analysis/HiggsAnalysis-CombinedLimit/blob/main/src/CascadeMinimizer.cc
+ def iterative_minimize(self, cascade:bool=True):
+ with Timer() as t:
+ min_tol = self.config['discrete_min_tol']
+ if self.min_nll is None:
+ self.min_nll = self.nll.getVal()
+ if abs(self.min_nll - self.nll.getVal()) > min_tol:
+ self.robust_minimize(cascade=cascade)
+
+ self.discrete_nuisance.freeze_discrete_params(True)
+
+ params_to_freeze = ROOT.RooArgSet(self.discrete_nuisance.all_params)
+ params_to_freeze.remove(self.discrete_nuisance.multipdf_params)
+ params_to_freeze.add(self.poi)
+ ROOT.RooStats.RemoveConstantParameters(params_to_freeze)
+ RooArgSet.set_constant_state(params_to_freeze, True)
+ nll_params = self.nll.getParameters(0)
+ nll_params.remove(self.discrete_nuisance.multipdf_cats)
+ ROOT.RooStats.RemoveConstantParameters(nll_params)
+ clean_snapshot = nll_params.snapshot()
+ # now cycle and fit
+ status = 0
+ # start frm simplest scan, this is the full scan if do_short_combination is off
+ self.multiple_minimize(clean_snapshot=clean_snapshot, cascade=cascade, mode=0)
+ RooArgSet.set_constant_state(params_to_freeze, False)
+ # run one last fully floating fit to maintain RooFitResult
+ status = self.robust_minimize(cascade=cascade)
+ self.discrete_nuisance.freeze_discrete_params(False)
+ self.stdout.info(f'Finished iteration. Total time taken: {t.interval:.3f} s.')
+ return status
+
+ # taken from https://github.com/cms-analysis/HiggsAnalysis-CombinedLimit/blob/main/src/CascadeMinimizer.cc
+ def multiple_minimize(self, clean_snapshot:"ROOT.RooArgSet", cascade:bool=True, mode:int=0,
+ contributing_indices:Optional[List[np.ndarray]]=None):
+ if not self.has_discrete_nuisance():
+ raise RuntimeError('multiple minimize should only be used when discrete nuisances are available')
+ hide_constants = self.config['multimin_hide_constants']
+ mask_constraints = self.config['multimin_mask_constraints']
+ mask_channels = self.config['multimin_mask_channels']
+ mask_channels_ex = (mask_channels == 2)
+ self.caching_nll.set_nll(self.nll)
+
+ strategy_tmp = self.config['strategy']
+ self.config['strategy'] = 0
+
+ new_discrete_minimum = False
+ multipdf_cats = self.discrete_nuisance.multipdf_cats
+ num_cats = len(multipdf_cats)
+ pdf_sizes = self.discrete_nuisance.get_pdf_sizes()
+ best_indices = self.discrete_nuisance.get_current_pdf_indices()
+ self.stdout.info(f'Current index combination: {list(best_indices)}')
+ self.stdout.info(f'Current NLL: {self.nll.getVal()}')
+
+ if (mode == 0) or (contributing_indices is None):
+ contributing_indices = self.discrete_nuisance.create_contributing_indices()
+
+ # keep hold of best fitted parameters
+ nll_params = self.nll.getParameters(0)
+ nll_params.remove(self.discrete_nuisance.multipdf_cats)
+ nll_params_snapshot = nll_params.snapshot()
+
+ if mask_channels:
+ self.caching_nll.set_mask_non_discrete_channels(True)
+ if hide_constants:
+ self.caching_nll.set_hide_constants(True)
+ if mask_constraints:
+ self.caching_nll.set_mask_constraints(True)
+ self.minimizer = None
+
+ if (mode in [0, 1]):
+ n_combination = self.discrete_nuisance.get_n_orthogonal_combination()
+ self.stdout.info(f'Generating {n_combination} orthogonal index combinations.')
+ combinations = self.discrete_nuisance.get_orthogonal_combinations()
+ else:
+ n_combination = self.discrete_nuisance.get_n_orthogonal_combination(contributing_indices)
+ self.stdout.info(f'Generating {n_combination} index combinations.')
+ combinations = self.discrete_nuisance.get_total_combinations(contributing_indices)
+
+ # reorder combinations starting from indices closest to the best indices
+ combinations = self.discrete_nuisance.reorder_combinations(combinations, best_indices)
+
+ # filter combinations that are not contributing
+ combinations = self.discrete_nuisance.filter_combinations(combinations, contributing_indices)
+
+ self.stdout.info(f'Total number of combinations after filtering contributing indices: {len(combinations)}')
+
+ new_discrete_minimum = False
+ # skip the best fit case if already done
+ i_start = 1 if (mode != 0) else 0
+ self.stdout.info(f'Begin fast loop minimization of {len(combinations)} index combinations.')
+ with Timer() as t:
+ fit_counter = 0
+ max_deviation = 5
+ multipdf_cats = self.discrete_nuisance.multipdf_cats
+ # the overhead in this for loop should be way less than the fit time
+ for combination in combinations[i_start:]:
+ changed_index = self.discrete_nuisance.set_category_indices(combination)
+ if fit_counter > 0:
+ nll_params.assignValueOnly(clean_snapshot)
+ if mask_channels_ex:
+ self.discrete_nuisance.fix_non_target_cats(changed_index)
+ self.caching_nll.set_mask_non_discrete_channels(True)
+ self.discrete_nuisance.freeze_discrete_params(True)
+ status = self.robust_minimize(cascade=cascade)
+ if mask_channels_ex:
+ self.discrete_nuisance.float_all_cats()
+ self.caching_nll.set_mask_non_discrete_channels(False)
+
+ self.discrete_nuisance.freeze_discrete_params(False)
+ fit_counter += 1
+ current_nll = self.nll.getVal()
+ delta_nll = current_nll - self.min_nll
+ self.stdout.debug(f'Index combination: {list(combination)}, NLL = {current_nll}, delta NLL = {delta_nll}')
+ # found new minimum
+ if (delta_nll < 0):
+ self.min_nll = current_nll
+ nll_params_snapshot.assignValueOnly(nll_params)
+ # set the best indices again
+ if not np.array_equal(best_indices, combination):
+ self.stdout.info(f'Found a better minimum at {list(combination)}. '
+ f'New NLL = {current_nll}, delta NLL = {delta_nll}.')
+ new_discrete_minimum = True
+ best_indices = combination.copy()
+ # discard pdf that gives large nll
+ if (mode == 1):
+ if (delta_nll > max_deviation):
+ index_diff = np.where(best_indices != combination)[0]
+ diff_count = index_diff.shape[0]
+ if diff_count == 1:
+ index_diff = index_diff[0]
+ cat = self.discrete_nuisance.multipdf_cats.at(index_diff)
+ cat_index = cat.getIndex()
+ if (cat_index != best_indices[index_diff]):
+ contributing_indices[index_diff][cat_index] = 0
+ pdf_name = self.discrete_nuisance.multipdfs.at(index_diff).GetName()
+ self.stdout.info(f'Found pdf index that gives large nll. Discarding pdf index '
+ f'{cat_index} from the multipdf "{pdf_name}"')
+ # assign best indices
+ self.discrete_nuisance.set_category_indices(best_indices)
+ nll_params.assignValueOnly(nll_params_snapshot)
+ n_combination_run = len(combinations) - i_start
+ time_per_combination = (t.interval / n_combination_run)
+
+ self.stdout.info(f'Done {n_combination_run} combinations in {t.interval:.3f} s '
+ f'({time_per_combination:.4f} s per combination). New discrete minimum? {new_discrete_minimum}')
+
+ self.config['strategy'] = strategy_tmp
+
+ if mask_channels:
+ self.caching_nll.set_mask_non_discrete_channels(False)
+
+ if hide_constants:
+ self.caching_nll.set_hide_constants(False)
+ if mask_constraints:
+ self.caching_nll.set_mask_constraints(False)
+ self.minimizer = None
+
+ return new_discrete_minimum
+
def use_limits(self, par:"ROOT.RooRealVar", val:float):
if (val < par.getMin()):
self.stdout.warning(f'ExtendedMinimizer::use_limits("{self.name}") {par.GetName()} = {val} '
@@ -761,3 +1173,54 @@ class ExtendedMinimizer(AbstractObject):
g_interpolated.SetMarkerStyle(20)
return (g, g_interpolated)
+
+ def eigen_analysis(self):
+ if not self.minimizer:
+ self.stdout.warning('Failed to get Hessian matrix: minimizer not set')
+ return None
+ # Obtain Hessian matrix either from patched Minuit or after inversion
+ # TMatrixDSym G = Minuit2::MnHesse::lastHessian();
+ last_fit = self.minimizer.lastMinuitFit()
+ if not last_fit:
+ self.stdout.warning('Failed to get Hessian matrix: no fit performed')
+ return None
+
+ self.Hessian_matrix = last_fit.covarianceMatrix().Invert()
+
+ if not isinstance(self.Hessian_matrix, ROOT.TMatrixDSym):
+ raise ValueError('invalid Hessian matrix')
+ n = self.Hessian_matrix.GetNrows()
+
+ # construct reduced Hessian matrix
+ Gred = ROOT.TMatrixDSym(n)
+ for i in range(n):
+ for j in range(n):
+ norm = math.sqrt(self.Hessian_matrix(i, i)*self.Hessian_matrix(j, j))
+ Gred[i][j] = self.Hessian_matrix(i, j)/norm
+
+ # perform eigenvalue analysis using ROOT standard tools
+ Geigen = ROOT.TMatrixDSymEigen(Gred)
+
+ self.eigen_values = Geigen.GetEigenValues()
+ self.eigen_vectors = Geigen.GetEigenVectors()
+
+ # simple printing of eigenvalues and eigenvectors
+ self.eigen_values.Print()
+ self.eigen_vectors.Print()
+
+ def get_floating_nll_params(self):
+ if not self.nll:
+ raise RuntimeError('NLL not initialized')
+ nll_params = self.nll.getParameters(0)
+ ROOT.RooStats.RemoveConstantParameters(nll_params)
+ return nll_params
+
+ def print_floating_nll_params(self):
+ nll_params = self.get_floating_nll_params()
+ for param in nll_params:
+ if param.InheritsFrom('RooRealVar'):
+ self.stdout.info(f'{param.GetName()} {param.getVal()}')
+ elif param.InheritsFrom('RooCategory'):
+ self.stdout.info(f'{param.GetName()} {param.currentIndex()}')
+ else:
+ self.stdout.info(f'{param.GetName()}')
\ No newline at end of file
diff --git a/quickstats/components/extended_model.py b/quickstats/components/extended_model.py
index 266acc26d2fb4872f4470f36cb739497c3dbb2ed..ffadcb4aae997260d1edbeaeb692e688df72054a 100644
--- a/quickstats/components/extended_model.py
+++ b/quickstats/components/extended_model.py
@@ -22,7 +22,7 @@ from quickstats import semistaticmethod, AbstractObject
from quickstats.core.io import switch_verbosity
from quickstats.maths.numerics import is_integer, pretty_value, get_bins_given_edges, array_issubset, get_rmin_rmax
from quickstats.parsers import RooParamSetupParser
-from quickstats.utils.root_utils import load_macro
+from quickstats.utils.root_utils import load_macro, close_read_cache
from quickstats.utils.common_utils import str_list_filter, combine_dict, filter_by_wildcards
from quickstats.utils import roofit_utils
from quickstats.utils.string_utils import split_str, remove_whitespace
@@ -31,6 +31,7 @@ from quickstats.interface.root import roofit_extension as rf_ext
from quickstats.interface.root.roofit_extension import get_str_data
from quickstats.components.basics import WSArgument, SetValueMode, ConstraintType
from .extended_minimizer import ExtendedMinimizer
+from .discrete_nuisance import DiscreteNuisance
class ExtendedModel(AbstractObject):
@@ -54,6 +55,20 @@ class ExtendedModel(AbstractObject):
'range_sideband_high': 'SBHi'
}
+ _DEFAULT_CONFIGS_ = {
+ 'filename': None,
+ 'ws_name': None,
+ 'mc_name': None,
+ 'data_name': None,
+ 'initial_snapshots': None,
+ 'do_discrete_nuisance': True,
+ 'do_fix_cache': True,
+ 'do_fix_multi': True,
+ 'do_binned_likelihood': True,
+ 'interpolation_code': -1,
+ 'tag_as_measurement': None,
+ }
+
_DEFAULT_CONTENT_ = {
WSArgument.PDF : kClassName|kName|kArgs,
WSArgument.FUNCTION : kClassName|kName|kArgs
@@ -63,23 +78,28 @@ class ExtendedModel(AbstractObject):
data_name:Optional[str]="combData", snapshot_name:Optional[Union[List[str], str]]=None,
binned_likelihood:bool=True, tag_as_measurement:Optional[str]=None,
fix_cache:bool=True, fix_multi:bool=True, interpolation_code:int=-1,
- load_extension:bool=True, minimizer_cls=None, verbosity:Optional[Union[int, str]]="INFO"):
- super().__init__(verbosity=verbosity)
- self.filename = filename
- self.ws_name = ws_name
- self.mc_name = mc_name
- self.data_name = data_name
- self.initial_snapshots = snapshot_name
- self.binned_likelihood = binned_likelihood
- self.tag_as_measurement = tag_as_measurement
- self.fix_cache = fix_cache
- self.fix_multi = fix_multi
- self.interpolation_code = interpolation_code
- self.last_fit_status = None
- if minimizer_cls is None:
- self.minimizer_cls = ExtendedMinimizer
- else:
- self.minimizer_cls = minimizer_cls
+ discrete_nuisance:bool=True, load_extension:bool=True, minimizer=None,
+ verbosity:Optional[Union[int, str]]="INFO"):
+ super().__init__(verbosity=verbosity)
+
+ self.set_minimizer(minimizer)
+
+ config = {
+ 'ws_name': ws_name,
+ 'mc_name': mc_name,
+ 'data_name': data_name,
+ 'initial_snapshots': snapshot_name,
+ 'do_discrete_nuisance': discrete_nuisance,
+ 'do_fix_cache': fix_cache,
+ 'do_fix_multi': fix_multi,
+ 'do_binned_likelihood': binned_likelihood,
+ 'interpolation_code': interpolation_code,
+ 'tag_as_measurement': tag_as_measurement
+ }
+
+ self.config = combine_dict(self._DEFAULT_CONFIGS_, config)
+ # avoid copying RooWorkspace instance
+ self.config['filename'] = filename
quickstats.load_corelib()
if load_extension:
@@ -118,25 +138,17 @@ class ExtendedModel(AbstractObject):
def observables(self):
return self._observables
@property
+ def discrete_nuisance(self):
+ return self._discrete_nuisance
+ @property
def category(self):
return self._category
@property
def floating_auxiliary_variables(self):
return self._floating_auxiliary_variables
@property
- def initial_snapshots(self):
- return self._initial_snapshots
-
- @initial_snapshots.setter
- def initial_snapshots(self, val):
- if val is None:
- self._initial_snapshots = []
- elif isinstance(val, str):
- self._initial_snapshots = split_str(val, sep=',', remove_empty=True)
- elif isinstance(val, list):
- self._initial_snapshots = val
- else:
- raise ValueError('"initial_snapshots" must be string or list of strings')
+ def minimizer(self):
+ return self._minimizer
@semistaticmethod
def load_extension(self):
@@ -194,59 +206,69 @@ class ExtendedModel(AbstractObject):
self.stdout.info(f'Deactivated level 2 constant term optimization for {name}')
def initialize(self):
- if isinstance(self.filename, str):
- if not os.path.exists(self.filename):
- raise FileNotFoundError(f'workspace file "{self.filename}" does not exist')
- self.stdout.info(f'Opening file "{self.filename}"')
- file = ROOT.TFile(self.filename)
+ filename = self.config['filename']
+ ws_name = self.config['ws_name']
+ mc_name = self.config['mc_name']
+ if isinstance(filename, str):
+ if not os.path.exists(filename):
+ raise FileNotFoundError(f'workspace file "{filename}" does not exist')
+ self.stdout.info(f'Opening file "{filename}"')
+ file = ROOT.TFile(filename)
if (not file):
- raise RuntimeError(f"Something went wrong while loading the root file: {self.filename}")
+ raise RuntimeError(f"Something went wrong while loading the root file: {filename}")
+ # remove read cache
+ close_read_cache(file)
# load workspace
- if self.ws_name is None:
+ if ws_name is None:
ws_names = [i.GetName() for i in file.GetListOfKeys() if i.GetClassName() == 'RooWorkspace']
if not ws_names:
- raise RuntimeError(f"No workspaces found in the root file: {self.filename}")
+ raise RuntimeError(f"No workspaces found in the root file: {filename}")
if len(ws_names) > 1:
self.stdout.warning("Found multiple workspace instances from the root file: "
f"{self.filename}. Available workspaces are \"{','.join(ws_names)}\". "
f"Will choose the first one by default")
- self.ws_name = ws_names[0]
- ws = file.Get(self.ws_name)
- elif isinstance(self.filename, ROOT.RooWorkspace):
+ ws_name = ws_names[0]
+ ws = file.Get(ws_name)
+ elif isinstance(filename, ROOT.RooWorkspace):
file = None
- ws = self.filename
+ ws = filename
if not ws:
- raise RuntimeError(f'failed to load workspace "{self.ws_name}"')
- self.ws_name = ws.GetName()
- self.stdout.info(f'Loaded workspace "{self.ws_name}"')
+ raise RuntimeError(f'failed to load workspace "{ws_name}"')
+ ws_name = ws.GetName()
+ self.config['ws_name'] = ws_name
+ self.stdout.info(f'Loaded workspace "{ws_name}"')
# load model config
- if self.mc_name is None:
+ if mc_name is None:
mc_names = [i.GetName() for i in ws.allGenericObjects() if 'ModelConfig' in i.ClassName()]
if not mc_names:
- raise RuntimeError(f"no ModelConfig object found in the workspace: {self.ws_name}")
+ raise RuntimeError(f"no ModelConfig instance found in the workspace \"{ws_name}\"")
if len(mc_names) > 1:
- self.stdout.warning(f"Found multiple ModelConfig instances from the workspace: {self.ws_name}. "
- f"Available ModelConfigs are \"{','.join(mc_names)}\". "
- "Will choose the first one by default")
- self.mc_name = mc_names[0]
- model_config = ws.obj(self.mc_name)
+ mc_names_str = ', '.join([f'"{mc_name}"' for mc_name in mc_names])
+ self.stdout.warning(f"Found multiple ModelConfig instances in the workspace \"{ws_name}\""
+ f"Available ModelConfigs are {mc_names_str}. "
+ "Will choose the first one by default.")
+ mc_name = mc_names[0]
+ model_config = ws.obj(mc_name)
+ self.config['mc_name'] = mc_name
if not model_config:
- raise RuntimeError(f'failed to load model config "{self.mc_name}"')
- self.stdout.info(f'Loaded model config "{self.mc_name}"')
+ raise RuntimeError(f'failed to load model config "{mc_name}"')
+ self.stdout.info(f'Loaded model config "{mc_name}"')
# modify interpolation code
- if self.interpolation_code != -1:
- self.modify_interp_codes(ws, self.interpolation_code,
+ interpolation_code = self.config['interpolation_code']
+ if interpolation_code != -1:
+ self.modify_interp_codes(ws, interpolation_code,
classes=[ROOT.RooStats.HistFactory.FlexibleInterpVar, ROOT.PiecewiseInterpolation])
# activate binned likelihood
- if self.binned_likelihood:
+ if self.config['do_binned_likelihood']:
self.activate_binned_likelihood(ws)
# set main measurement
- if self.tag_as_measurement:
- self.set_measurement(ws, condition=lambda name: name.startswith(self.tag_as_measurement))
+ tag_as_measurement = self.config['tag_as_measurement']
+ if tag_as_measurement:
+ self.set_measurement(ws, condition=lambda name: name.startswith(tag_as_measurement))
# deactivate level 2 constant term optimization
self.deactivate_lv2_const_optimization(ws,
@@ -259,19 +281,22 @@ class ExtendedModel(AbstractObject):
self.stdout.info(f'Loaded model pdf "{pdf.GetName()}" from model config')
# load dataset
- if self.data_name is None:
+ data_name = self.config['data_name']
+ if data_name is None:
data_names = [i.GetName() for i in ws.allData()]
if not data_names:
- raise RuntimeError(f"no datasets found in the workspace: {ws.GetName()}")
- self.data_name = data_names[0]
- data = ws.data(self.data_name)
+ raise RuntimeError(f"no datasets found in the workspace: {ws_name}")
+ data_name = data_names[0]
+ data = ws.data(data_name)
# in case there is a bug in hash map
if not data:
- data = [i for i in ws.allData() if i.GetName() == self.data_name]
+ data = [i for i in ws.allData() if i.GetName() == data_name]
if not data:
- raise RuntimeError(f'failed to load dataset "{self.data_name}"')
+ raise RuntimeError(f'failed to load dataset "{data_name}"')
data = data[0]
- self.stdout.info(f'Loaded dataset "{data.GetName()}" from workspace')
+ data_name = data.GetName()
+ self.config['data_name'] = data_name
+ self.stdout.info(f'Loaded dataset "{data_name}" from workspace')
# load nuisance parameters
nuisance_parameters = model_config.GetNuisanceParameters()
@@ -315,6 +340,18 @@ class ExtendedModel(AbstractObject):
else:
category = None
category_pdfs = None
+
+ # check channel masks (for RooSimultaneousOpt)
+ if (pdf.InheritsFrom("RooSimultaneousOpt") or hasattr(pdf, 'channelMasks')):
+ channel_masks = pdf.channelMasks()
+ n_channel_masks = len(channel_masks)
+ if (n_channel_masks > 0):
+ active_masks = [m for m in channel_masks if m.getVal() > 0]
+ n_active_masks = len(active_masks)
+ self.stdout.info(f'{n_active_masks} out of {n_channel_masks} channels masked')
+ if self.stdout.verbosity <= "INFO":
+ self.stdout.info('Channel masks:')
+ channel_masks.Print()
self._file = file
self._workspace = ws
@@ -328,12 +365,34 @@ class ExtendedModel(AbstractObject):
self._observables = observables
self._category = category
self._floating_auxiliary_variables = None
-
- # Load snapshots
- self.load_snapshots(self.initial_snapshots)
+
+ self.last_fit_status = None
+
+ # parse initial snapshots
+ initial_snapshots = self.config['initial_snapshots']
+ if initial_snapshots is None:
+ initial_snapshots = []
+ elif isinstance(initial_snapshots, str):
+ initial_snapshots = split_str(initial_snapshots, sep=',', remove_empty=True)
+ elif isinstance(initial_snapshots, list):
+ initial_snapshots = list(initial_snapshots)
+ else:
+ raise ValueError('"initial_snapshots" must be string or list of strings')
+ self.config['initial_snapshots'] = initial_snapshots
+ # load snapshots
+ self.load_snapshots(initial_snapshots)
+
+ # discrete nuisance
+ if self.config['do_discrete_nuisance']:
+ self._discrete_nuisance = DiscreteNuisance(ws, pdf, verbosity=self.stdout.verbosity)
+ else:
+ self._discrete_nuisance = DiscreteNuisance(verbosity=self.stdout.verbosity)
RooMsgService.remove_topics()
return None
+
+ def set_minimizer(self, minimizer:Optional["ExtendedMinimizer"]=None):
+ self._minimizer = minimizer
def set_parameters(self, param_setup: Optional[Union[str, Sequence, Dict]] = None,
params: Optional["ROOT.RooArgSet"] = None,
@@ -491,6 +550,9 @@ class ExtendedModel(AbstractObject):
raise ValueError(f'Object "{name}" not found in the workspace')
return obj
+ def has_discrete_nuisance(self):
+ return self.discrete_nuisance.has_discrete_nuisance()
+
@staticmethod
def randomize_globs(pdf:ROOT.RooAbsPdf, globs:ROOT.RooArgSet, seed:int):
"""Randomize values of global observables (for generating pseudo-experiments)
@@ -1003,24 +1065,30 @@ class ExtendedModel(AbstractObject):
if do_fit:
if dataset is None:
dataset = self.data
- minimizer = self.minimizer_cls("Minimizer", self.pdf, dataset, workspace=self.workspace)
- if minimizer_options is None:
- minimizer_options = combine_dict(minimizer._DEFAULT_MINIMIZER_OPTION_)
- if nll_options is None:
- nll_options = combine_dict(minimizer._DEFAULT_NLL_OPTION_)
- minimizer.configure(**minimizer_options)
- if constraint_option == 0:
- pass
- elif constraint_option == 1:
- self.set_constrained_nuisance_parameters_to_nominal()
- else:
- raise ValueError(f"unsupported constraint option: {constraint_option}")
- if isinstance(nll_options, dict):
- minimizer.configure_nll(**nll_options)
- elif isinstance(nll_options, list):
- minimizer.set_nll_commands(nll_options)
+ # create new instance of minimizer
+ if self.minimizer is None:
+ minimizer = self.minimizer_cls("Minimizer", self.pdf, dataset, workspace=self.workspace)
+ if minimizer_options is None:
+ minimizer_options = combine_dict(minimizer._DEFAULT_MINIMIZER_OPTION_)
+ if nll_options is None:
+ nll_options = combine_dict(minimizer._DEFAULT_NLL_OPTION_)
+ minimizer.configure(**minimizer_options)
+ if constraint_option == 0:
+ pass
+ elif constraint_option == 1:
+ self.set_constrained_nuisance_parameters_to_nominal()
+ else:
+ raise ValueError(f"unsupported constraint option: {constraint_option}")
+ if isinstance(nll_options, dict):
+ minimizer.configure_nll(**nll_options)
+ elif isinstance(nll_options, list):
+ minimizer.set_nll_commands(nll_options)
+ else:
+ raise ValueError(f"unsupported nll options format")
else:
- raise ValueError(f"unsupported nll options format")
+ self.minimizer.set_pdf(self.pdf)
+ self.minimizer.set_data(dataset)
+ minimizer = self.minimizer
if poi_set is not None:
uncond_fit = poi_profile is None
@@ -1071,7 +1139,7 @@ class ExtendedModel(AbstractObject):
mu_cond=format_mu_str(poi_profile))
else:
asimov_data_name = names['asimov_no_poi']
-
+
if method == "old":
channel_asimov_data_name = names['channel_asimov']
sim_pdf = self.pdf
@@ -1285,15 +1353,18 @@ class ExtendedModel(AbstractObject):
cat_label = cat_data.first
cat_index = cat_data.second
pdf_cat = pdf.getPdf(cat_label)
- obs = pdf_cat.getObservables(self.observables)
- target_obs = obs.first()
+ observables = pdf_cat.getObservables(self.observables)
category_map[cat_label] = {}
category_map[cat_label]['index'] = cat_index
category_map[cat_label]['pdf'] = pdf_cat.GetName()
- category_map[cat_label]['observable'] = target_obs.GetName()
- bin_range = target_obs.getRange()
- category_map[cat_label]['bin_range'] = (bin_range.first, bin_range.second)
- category_map[cat_label]['bins'] = target_obs.getBins()
+ category_map[cat_label]['observables'] = {}
+ for observable in observables:
+ obs_name = observable.GetName()
+ bin_range = observable.getRange()
+ category_map[cat_label]['observables'][obs_name] = {
+ 'bin_range': (bin_range.first, bin_range.second),
+ 'bins': observable.getBins()
+ }
return category_map
def get_categories(self):
@@ -1303,10 +1374,13 @@ class ExtendedModel(AbstractObject):
category_map = self.get_category_map()
binnings = {}
for category in category_map:
+ if len(category_map[category]['observables']) > 1:
+ raise RuntimeError('rebinning of multi-observable category is not supported')
+ orig_binnings = next(iter(category_map[category]['observables'].values()))
binnings[category] = {}
- bin_range = category_map[category]['bin_range']
+ bin_range = orig_binnings['bin_range']
if bins is None:
- _bins = category_map[category]['bins']
+ _bins = orig_binnings['bins']
elif isinstance(bins, dict):
_bins = bins.get(category, None)
if _bins is None:
@@ -1332,7 +1406,11 @@ class ExtendedModel(AbstractObject):
is_binned = {}
for cat, cat_data in category_map.items():
distributions[cat] = {}
- observable = category_map[cat]['observable']
+ observables = category_map[cat]['observables']
+ if len(observables) > 1:
+ raise RuntimeError('found multi-observable category')
+ observable = next(iter(observables))
+ observable_binnings = next(iter(observables.values()))
if observable not in dataset_values:
raise RuntimeError(f"no data associated with the observable \"{observable}\" found in the dataset")
distributions[cat]['observable'] = observable
@@ -1342,7 +1420,7 @@ class ExtendedModel(AbstractObject):
ind = np.argsort(x)
x = x[ind]
y = dataset_values['weight'][mask][ind]
- default_bins = category_map[cat]['bins']
+ default_bins = observable_binnings['bins']
ghost = False
if np.all(y == 1.):
binned = False
@@ -1633,7 +1711,10 @@ class ExtendedModel(AbstractObject):
binnings['all_cat'] = combine_dict(binnings[categories[0]])
for i, category in enumerate(categories):
- obs_name = category_map[category]['observable']
+ observables = category_map[category]['observables']
+ if len(observables) > 1:
+ raise RuntimeError('multi-observable category is not supported')
+ obs_name = next(iter(observables))
if discriminant is None:
xlabel = obs_name
else:
@@ -1694,6 +1775,8 @@ class ExtendedModel(AbstractObject):
raise ValueError(f"the collection \"{variable_type}\" does not contain members of the type \"RooRealVar\"")
if resolved_vtype == WSArgument.VARIABLE:
variables = ROOT.RooArgSet(self.workspace.allVars())
+ if self.has_discrete_nuisance():
+ variables.add(self.discrete_nuisance.multipdf_cats)
elif resolved_vtype == WSArgument.OBSERVABLE:
if isinstance(self.pdf, ROOT.RooSimultaneous):
variables = self.observables.Clone()
@@ -1738,6 +1821,8 @@ class ExtendedModel(AbstractObject):
if self.floating_auxiliary_variables is None:
self._load_floating_auxiliary_variables()
variables.add(self.floating_auxiliary_variables)
+ if self.has_discrete_nuisance():
+ variables.add(self.discrete_nuisance.multipdf_cats)
else:
raise ValueError(f"unknown variable type \"{variable_type}\"")
if sort:
@@ -1801,11 +1886,30 @@ class ExtendedModel(AbstractObject):
@staticmethod
def _format_category_summary(category_name:str, category_map:Dict):
- observable = category_map['observable']
- bin_range = category_map['bin_range']
- rmin, rmax = get_rmin_rmax(bin_range)
- bins = category_map['bins']
- summary_str = f"{category_name} (observable = {observable}, range = [{rmin}, {rmax}], bins = {bins})"
+ all_names = []
+ all_ranges = []
+ all_bins = []
+ for obs_name in category_map['observables']:
+ all_names.append(obs_name)
+ bin_range = category_map['observables'][obs_name]['bin_range']
+ rmin, rmax = get_rmin_rmax(bin_range)
+ all_ranges.append(f'[{rmin}, {rmax}]')
+ bins = category_map['observables'][obs_name]['bins']
+ all_bins.append(f'{bins}')
+ if len(all_names) == 1:
+ summary_str = f"{category_name} (observable = {all_names[0]}, range = {all_ranges[0]}, bins = {all_bins[0]})"
+ else:
+ summary_str = (f"{category_name} (observables = {'(' + ', '.join(all_names) + ')'}, "
+ f"ranges = {'(' + ', '.join(all_ranges) + ')'}, "
+ f"bins = {'(' + ', '.join(all_bins) + ')'})")
+ return summary_str
+
+ @staticmethod
+ def _format_multipdf_summary(multipdf, multipdf_cat):
+ multipdf_name = multipdf.GetName()
+ category_name = multipdf_cat.GetName()
+ num_cat = len(multipdf_cat)
+ summary_str = f"{multipdf_name} (index_cat = {category_name}, num_pdfs = {num_cat})"
return summary_str
def print_summary(self, items:Optional[List]=None, suppress_print:bool=False, detailed:bool=True,
@@ -1813,7 +1917,7 @@ class ExtendedModel(AbstractObject):
save_as:Optional[str]=None):
if items is None:
items = ['workspace', 'dataset', 'snapshot', 'category',
- 'poi', 'detailed_nuisance_parameter']
+ 'poi', 'detailed_nuisance_parameter', 'multipdf']
summary_str = ""
# workspace
if 'workspace' in items:
@@ -1840,6 +1944,14 @@ class ExtendedModel(AbstractObject):
summary_str += f"Categories ({n_cat}):\n"
for category in category_map:
summary_str += "\t" + self._format_category_summary(category, category_map[category]) + "\n"
+ # multi pdfs
+ if 'multipdf' in items:
+ multipdf_cats = self.discrete_nuisance.multipdf_cats
+ multipdfs = self.discrete_nuisance.multipdfs
+ n_multipdf = len(multipdfs)
+ summary_str += f"MultiPdfs ({n_multipdf}):\n"
+ for multipdf, multipdf_cat in zip(multipdfs, multipdf_cats):
+ summary_str += "\t" + self._format_multipdf_summary(multipdf, multipdf_cat) + "\n"
# pois, NPs
param_strs = []
param_sets = []
@@ -1971,4 +2083,9 @@ class ExtendedModel(AbstractObject):
if isinstance(variables, (str, WSArgument)):
variables = self.get_variables(variables)
self.workspace.saveSnapshot(snapshot_name, variables)
- self.stdout.info(f'Saved snapshot "{snapshot_name}"')
\ No newline at end of file
+ self.stdout.info(f'Saved snapshot "{snapshot_name}"')
+
+
+ def todo(self):
+ # remove NPs from fall back discrete pdf
+ pass
\ No newline at end of file
diff --git a/quickstats/components/likelihood.py b/quickstats/components/likelihood.py
index 91e7b5af85680d45d75fa9ce2da70efc50869b9a..45db367bfaffd4f7823eed8a08c4fc643cf06d64 100644
--- a/quickstats/components/likelihood.py
+++ b/quickstats/components/likelihood.py
@@ -12,7 +12,7 @@ from quickstats import DescriptiveEnum
from quickstats.maths.numerics import pretty_value
from quickstats.components.basics import WSArgument
from quickstats.components import AnalysisObject
-from quickstats.utils.common_utils import parse_config
+from quickstats.utils.common_utils import parse_config, combine_dict
from quickstats.parsers import RooParamSetupParser
class FitMode(DescriptiveEnum):
@@ -24,11 +24,14 @@ class FitMode(DescriptiveEnum):
class Likelihood(AnalysisObject):
- def __init__(self, filename:str, poi_name:Optional[Union[str, List[str]]]=None,
+ def __init__(self, filename:str,
+ poi_name:Optional[Union[str, List[str]]]=None,
data_name:str='combData',
config:Optional[Dict]=None,
- verbosity:Optional[Union[int, str]]="INFO"):
+ verbosity:Optional[Union[int, str]]="INFO",
+ **kwargs):
config = parse_config(config)
+ config = combine_dict(config, kwargs)
config['filename'] = filename
config['poi_name'] = poi_name
config['data_name'] = data_name
@@ -73,6 +76,10 @@ class Likelihood(AnalysisObject):
else:
combined_fit_result['significance'] = None
combined_fit_result['pvalue'] = None
+ if (uncond_fit_result['status'] != 0) or (cond_fit_result['status'] != 0):
+ combined_fit_result['status'] = -1
+ else:
+ combined_fit_result['status'] = 0
return combined_fit_result
def nll_fit(self, poi_val:Optional[Union[Dict[str, float], float]]=None,
diff --git a/quickstats/components/workspaces/ws_decomposer.py b/quickstats/components/workspaces/ws_decomposer.py
index 68b7c03a79e42e55ebf8f6fc2f614a18dd3b557a..c95b7aa357db84e5a38b0a566714c036b8ec1cb5 100644
--- a/quickstats/components/workspaces/ws_decomposer.py
+++ b/quickstats/components/workspaces/ws_decomposer.py
@@ -100,7 +100,6 @@ class WSDecomposer(XMLWSBase):
with Timer() as t:
model = ExtendedModel(infile, data_name=None, verbosity="WARNING")
orig_cat = model.pdf.indexCat()
- category_map = model.get_category_map()
categories = self.parse_categories(category_expr, orig_cat)
snapshots = self.parse_snapshots(model.workspace, snapshots_to_save)
if not categories:
@@ -129,7 +128,7 @@ class WSDecomposer(XMLWSBase):
pois_cat = param_cat.selectCommon(model.pois)
globs_cat = param_cat.selectCommon(model.global_observables)
if rebuild_nuis:
- nuis_cat = ROOT.RFUtils.SelectConstantArgs(param_cat, False)
+ nuis_cat = ROOT.RFUtils.GetConstantParameters(param_cat, False)
nuis_cat.remove(pois_cat, False, True)
nuis_cat.remove(globs_cat, False, True)
else:
diff --git a/quickstats/components/workspaces/xml_ws_combiner.py b/quickstats/components/workspaces/xml_ws_combiner.py
index e1c76661765df54f524b41510e48b9189e544b25..5a630ceaeb34124850200f564f2cbcb9ad0e203f 100644
--- a/quickstats/components/workspaces/xml_ws_combiner.py
+++ b/quickstats/components/workspaces/xml_ws_combiner.py
@@ -300,7 +300,7 @@ class XMLWSCombiner(XMLWSBase):
channel_config = self.channel_config[channel]
model = self._get_channel_model(channel_config)
strict_mode = self.combination_config["strict"]
- ws_filename = model.filename
+ ws_filename = model.config['filename']
renamed_pdfs = list(channel_config["rename_map"]["pdf"])
renamed_vars = list(channel_config["rename_map"]["var"])
renamed_pois = list(channel_config["poi_map"].values())
@@ -317,8 +317,8 @@ class XMLWSCombiner(XMLWSBase):
self._exception_check(msg, strict_mode)
continue
if not isinstance(pdf, ROOT.RooGaussian):
- msg = f"constraint PDF \"{pdf_name}\" that is not an instance of RooGaussian is not supported"
- self._exception_check(msg, strict_mode)
+ #msg = f"constraint PDF \"{pdf_name}\" that is not an instance of RooGaussian is not supported"
+ #self._exception_check(msg, strict_mode)
continue
nuis = model.workspace.var(nuis_name)
if not nuis:
diff --git a/quickstats/concurrent/__init__.py b/quickstats/concurrent/__init__.py
index 0fee6f6ad0fc8ef53b64dd9fbe448e89279ea398..734a345171393363079213d3e81116a02ee26415 100644
--- a/quickstats/concurrent/__init__.py
+++ b/quickstats/concurrent/__init__.py
@@ -2,4 +2,5 @@ from .abstract_runner import AbstractRunner
from .parameterised_runner import ParameterisedRunner
from .parameterised_asymptotic_cls import ParameterisedAsymptoticCLs
from .parameterised_likelihood import ParameterisedLikelihood
+from .parameterised_significance import ParameterisedSignificance
from .nuisance_parameter_ranking_runner import NuisanceParameterRankingRunner
\ No newline at end of file
diff --git a/quickstats/concurrent/abstract_runner.py b/quickstats/concurrent/abstract_runner.py
index ca84390c6750ffd994781674f6b769fca2df6817..494693f597aabe654f90a354ca064ac107f2e334 100644
--- a/quickstats/concurrent/abstract_runner.py
+++ b/quickstats/concurrent/abstract_runner.py
@@ -10,7 +10,7 @@ class AbstractRunner(AbstractObject):
def config(self):
return self._config
- def __init__(self, parallel:int=-1, timed:bool=True,
+ def __init__(self, parallel:int=-1,
save_log:bool=True, cache:bool=True,
verbosity:Optional[Union[int, str]]="INFO"):
super().__init__(verbosity=verbosity)
@@ -18,8 +18,7 @@ class AbstractRunner(AbstractObject):
self._config = {
'cache': cache,
'parallel': parallel,
- 'save_log': save_log,
- 'timed': timed
+ 'save_log': save_log
}
def _prerun_batch(self):
@@ -39,17 +38,19 @@ class AbstractRunner(AbstractObject):
def _end_of_instance_cleanup(self):
pass
-
+
+ @semistaticmethod
def get_instance_outpath(self, kwargs:Dict):
outpath = kwargs.get("outname", None)
return outpath
-
+
+ @semistaticmethod
def get_instance_logpath(self, kwargs:Dict):
outpath = self.get_instance_outpath(kwargs)
if outpath:
return os.path.splitext(outpath)[0] + ".log"
return None
-
+
def _is_valid_cache(self, cached_result):
return True
@@ -68,7 +69,7 @@ class AbstractRunner(AbstractObject):
pass
logpath = self.get_instance_logpath(kwargs)
- if logpath and self.config['save_log']:
+ if (logpath and self.config['save_log']) and (self.stdout.verbosity != 'DEBUG'):
with standard_log(logpath) as logger:
result = self._run_instance(**kwargs)
else:
@@ -84,8 +85,7 @@ class AbstractRunner(AbstractObject):
self._prerun_batch()
raw_result = execute_multi_tasks(self.run_instance, argument_list, parallel=parallel)
results = self.postprocess(raw_result, auxiliary_args)
- if self.config['timed']:
- self.stdout.info(f'All jobs have finished. Total time taken: {t.interval:.3f} s')
+ self.stdout.info(f'All jobs have finished. Total time taken: {t.interval:.3f} s')
return results
def run(self):
diff --git a/quickstats/concurrent/parameterised_asymptotic_cls.py b/quickstats/concurrent/parameterised_asymptotic_cls.py
index 2b36a6cac6f2173a1758a2762eb59677c0149e45..ef7d1cd1e80e5c3b774ea61b3a2091e9f9ae40cd 100644
--- a/quickstats/concurrent/parameterised_asymptotic_cls.py
+++ b/quickstats/concurrent/parameterised_asymptotic_cls.py
@@ -1,6 +1,5 @@
import os
import sys
-import copy
import json
from typing import Optional, Union, Dict, List, Any
from itertools import repeat
@@ -8,7 +7,7 @@ from itertools import repeat
from quickstats import semistaticmethod
from quickstats.parsers import ParamParser
from quickstats.concurrent import ParameterisedRunner
-from quickstats.utils.common_utils import batch_makedirs, json_load
+from quickstats.utils.common_utils import batch_makedirs, json_load, combine_dict, save_as_json
from quickstats.components import AsymptoticCLs
class ParameterisedAsymptoticCLs(ParameterisedRunner):
@@ -22,22 +21,16 @@ class ParameterisedAsymptoticCLs(ParameterisedRunner):
super().__init__(file_expr=file_expr, param_expr=param_expr,
filter_expr=filter_expr, exclude_expr=exclude_expr,
- parallel=parallel, timed=True, save_log=save_log,
+ parallel=parallel, save_log=save_log,
+ outdir=outdir, cachedir=cachedir,
cache=cache, allow_none=False, verbosity=verbosity)
- self.attributes = {
+ self.attributes.update({
'input_path': input_path,
'config': config,
- 'outdir': outdir,
- 'cachedir': cachedir,
'outname': outname,
'save_summary': save_summary
- }
-
- def _prerun_batch(self):
- outdir = self.attributes['outdir']
- cache_dir = self.get_cache_dir()
- batch_makedirs([outdir, cache_dir])
+ })
@semistaticmethod
def _prerun_instance(self, filename:str, parameters:Optional[Dict[str, Any]]=None, **kwargs):
@@ -46,13 +39,7 @@ class ParameterisedAsymptoticCLs(ParameterisedRunner):
else:
param_str = ""
self.stdout.info(f"Evaluating limit for the workspace {filename} {param_str}")
-
- @semistaticmethod
- def _cached_return(self, outname:str):
- with open(outname, 'r') as f:
- limits = json_load(f)
- return limits
-
+
@semistaticmethod
def _run_instance(self, filename:str, config:Dict[str, Any],
parameters:Dict[str, Any],
@@ -61,6 +48,7 @@ class ParameterisedAsymptoticCLs(ParameterisedRunner):
**kwargs):
try:
config['filename'] = filename
+ config.update(kwargs)
asymptotic_cls = AsymptoticCLs(**config)
asymptotic_cls.evaluate_limits()
if outname is not None:
@@ -70,9 +58,6 @@ class ParameterisedAsymptoticCLs(ParameterisedRunner):
sys.stdout.write(f"{e}\n")
return {}
- def get_cache_dir(self):
- return os.path.join(self.attributes['outdir'], self.attributes['cachedir'])
-
def prepare_task_inputs(self):
input_path = self.attributes['input_path']
cache_dir = self.get_cache_dir()
@@ -82,20 +67,12 @@ class ParameterisedAsymptoticCLs(ParameterisedRunner):
configs = []
base_config = self.attributes['config']
- if base_config is None:
- base_config = {}
for param_point in param_points:
- new_config = copy.deepcopy(base_config)
- int_params = param_point['internal_parameters']
- val_expr = ParamParser.val_encode_parameters(int_params)
- base_fix_param = new_config.get("fix_param", None)
- fix_param = []
- if base_fix_param:
- fix_param.append(base_fix_param)
- if val_expr:
- fix_param.append(val_expr)
- new_config['fix_param'] = ",".join(fix_param)
- configs.append(new_config)
+ base_fix_params = base_config.get("fix_param", None)
+ internal_params = param_point['internal_parameters']
+ fix_param_expr = self.join_param_setup(base_fix_params, internal_params)
+ config = combine_dict(base_config, {"fix_param": fix_param_expr})
+ configs.append(config)
param_dpd_kwargs = {
'filename': param_data['filenames'],
'config': configs,
@@ -103,7 +80,8 @@ class ParameterisedAsymptoticCLs(ParameterisedRunner):
'parameters': param_data['parameters']
}
param_ind_kwargs = {
- 'save_summary': self.attributes['save_summary']
+ 'save_summary' : self.attributes['save_summary'],
+ 'verbosity' : self.attributes['verbosity']
}
self.set_param_ind_kwargs(**param_ind_kwargs)
self.set_param_dpd_kwargs(**param_dpd_kwargs)
@@ -132,12 +110,11 @@ class ParameterisedAsymptoticCLs(ParameterisedRunner):
final_result.append({**params, **limit})
import pandas as pd
final_result = pd.DataFrame(final_result).to_dict('list')
-
+
+ poi_name = self.attributes['config'].get('poi_name', None)
outdir = self.attributes['outdir']
- outname = self.attributes['outname']
+ outname = self.attributes['outname'].format(poi_name=poi_name)
if outname is not None:
outpath = os.path.join(outdir, outname)
- with open(outpath, "w") as file:
- json.dump(final_result, file, indent=2)
- file.truncate()
\ No newline at end of file
+ save_as_json(final_result, outpath)
\ No newline at end of file
diff --git a/quickstats/concurrent/parameterised_likelihood.py b/quickstats/concurrent/parameterised_likelihood.py
index 7b449b81a5bb31f8c8988e1f726f25ea098c5b88..9ea52d6b33e2537866d418bfe5caf24551d663f2 100644
--- a/quickstats/concurrent/parameterised_likelihood.py
+++ b/quickstats/concurrent/parameterised_likelihood.py
@@ -10,12 +10,12 @@ import ROOT
from quickstats import semistaticmethod
from quickstats.parsers import ParamParser
from quickstats.concurrent import ParameterisedRunner
-from quickstats.utils.common_utils import batch_makedirs
-from quickstats.maths.numerics import pretty_value
+from quickstats.utils.common_utils import batch_makedirs, save_as_json
from quickstats.components import Likelihood
class ParameterisedLikelihood(ParameterisedRunner):
- def __init__(self, input_file:str, param_expr:str,
+
+ def __init__(self, input_path:str, param_expr:str, file_expr:Optional[str]=None,
filter_expr:Optional[str]=None, exclude_expr:Optional[str]=None,
data_name:str="combData", uncond_snapshot:Optional[str]=None,
config:Optional[Dict]=None, outdir:str="output", cachedir:str="cache",
@@ -23,20 +23,19 @@ class ParameterisedLikelihood(ParameterisedRunner):
save_log:bool=True, parallel:int=-1, allow_nan:bool=True,
verbosity:Optional[Union[int, str]]="INFO"):
- super().__init__(file_expr=None, param_expr=param_expr,
+ super().__init__(file_expr=file_expr, param_expr=param_expr,
filter_expr=filter_expr, exclude_expr=exclude_expr,
- parallel=parallel, timed=True, save_log=save_log,
+ parallel=parallel, save_log=save_log,
+ outdir=outdir, cachedir=cachedir,
cache=cache, allow_none=True, verbosity=verbosity)
-
- self.attributes = {
- 'input_file': input_file,
+
+ self.attributes.update({
+ 'input_path': input_path,
'data_name': data_name,
'uncond_snapshot': uncond_snapshot,
'config': config,
- 'outdir': outdir,
- 'cachedir': cachedir,
'outname': outname
- }
+ })
self.allow_nan = allow_nan
@@ -45,9 +44,7 @@ class ParameterisedLikelihood(ParameterisedRunner):
def _prerun_batch(self):
self.stdout.tips("When running likelihood scan on an Asimov dataset, remember to restore the global "
"observables to their Asimov values by loading the appropriate snapshot.")
- outdir = self.attributes['outdir']
- cache_dir = self.get_cache_dir()
- batch_makedirs([outdir, cache_dir])
+ super()._prerun_batch()
@semistaticmethod
def _prerun_instance(self, filename:str, mode:int, poi_val:Optional[Union[float, Dict[str, float]]]=None, **kwargs):
@@ -57,13 +54,6 @@ class ParameterisedLikelihood(ParameterisedRunner):
elif mode == 1:
self.stdout.info(f"Evaluating unconditional NLL for the workspace {filename}")
- @semistaticmethod
- def _cached_return(self, outname:str):
- with open(outname, 'r') as f:
- result = json.load(f)
- processed_result = self._process_result(result)
- return processed_result
-
def _is_valid_cache(self, cached_result):
if (not self.allow_nan) and math.isnan(cached_result['nll']):
self.stdout.info('Found cached result with nan nll. Retrying')
@@ -107,10 +97,10 @@ class ParameterisedLikelihood(ParameterisedRunner):
raise RuntimeError(error_msg)
if config is None:
config = {}
- verbosity = config.pop("verbosity", "INFO")
+ config.update(kwargs)
do_minos = config.pop("do_minos", False)
likelihood = Likelihood(filename=filename, poi_name=poi_name, data_name=data_name,
- config=config, verbosity=verbosity)
+ **config)
fit_result = likelihood.nll_fit(poi_val, mode=mode, do_minos=do_minos,
snapshot_name=snapshot_name)
# save results
@@ -124,9 +114,6 @@ class ParameterisedLikelihood(ParameterisedRunner):
except Exception as e:
sys.stdout.write(f"{e}\n")
return None
-
- def get_cache_dir(self):
- return os.path.join(self.attributes['outdir'], self.attributes['cachedir'])
def prepare_task_inputs(self):
@@ -134,47 +121,61 @@ class ParameterisedLikelihood(ParameterisedRunner):
if len(poi_names) == 0:
raise RuntimeError("no POI(s) to scan for")
- input_file = self.attributes['input_file']
+ input_path = self.attributes['input_path']
cache_dir = self.get_cache_dir()
outname = "{param_str}.json"
- param_points = self.get_param_points(input_file)
+ param_points = self.get_param_points(input_path)
+ n_param_points = len(param_points)
param_data = self.get_serialised_param_data(param_points, outdir=cache_dir, outname=outname)
-
+
+ filenames = param_data['filenames']
+ if not filenames:
+ raise RuntimeError(f"no input file found matching the expression: {input_path}")
+
+ external_param_points = self.get_external_param_points(input_path)
+ unique_filenames = []
+ outnames = []
+ poi_names_text = "_".join(poi_names)
+ for param_point in external_param_points:
+ unique_filenames.append(param_point['filename'])
+ param_str = ParamParser.str_encode_parameters(param_point['parameters'])
+ if param_str:
+ basename = f"{param_str}_{poi_names_text}_uncond"
+ else:
+ basename = f"{poi_names_text}_uncond"
+ outnames.append(os.path.join(cache_dir, basename))
+
+ n_unique_files = len(unique_filenames)
+
if self.attributes['uncond_snapshot'] is None:
# None is for unconditional NLL
- poi_values = [None]
+ poi_values = [None] * n_unique_files
# 1 is for unconditional NLL, 2 is for conditional NLL
- modes = [1] + [2]*len(param_points)
- snapshot_names = [Likelihood.kCurrentSnapshotName] * (1 + len(param_points))
+ modes = [1] * n_unique_files + [2] * n_param_points
+ snapshot_names = [Likelihood.kCurrentSnapshotName] * (n_unique_files + n_param_points)
else:
- poi_values = [None]
- modes = [4] + [2]*len(param_points)
+ poi_values = [None] * n_unique_files
+ modes = [4] * n_unique_files + [2] * n_param_points
uncond_snapshot_name = self.attributes['uncond_snapshot']
- snapshot_names = [uncond_snapshot_name] + [Likelihood.kCurrentSnapshotName] * len(param_points)
+ snapshot_names = ([uncond_snapshot_name] * n_unique_files +
+ [Likelihood.kCurrentSnapshotName] * n_param_points)
for param_point in param_points:
poi_values.append(param_point['internal_parameters'])
- outname_uncond = os.path.join(cache_dir, "{}_uncond.json".format("_".join(poi_names)))
param_dpd_kwargs = {
- 'poi_val': poi_values,
- 'mode': modes,
- 'snapshot_name': snapshot_names,
- 'outname': [outname_uncond] + param_data['outnames']
+ 'filename' : unique_filenames + filenames,
+ 'poi_val' : poi_values,
+ 'mode' : modes,
+ 'snapshot_name' : snapshot_names,
+ 'outname' : outnames + param_data['outnames']
}
- filename = list(set(param_data['filenames']))
-
- if len(filename) == 0:
- raise RuntimeError(f"no input file found matching the expression: {input_file}")
- if len(filename) > 1:
- raise RuntimeError("multiple input files detected: {}".format(", ".join(filename)))
-
param_ind_kwargs = {
- 'filename': filename[0],
- 'poi_name': self.attributes['poi_names'],
- 'data_name': self.attributes['data_name'],
- 'config': self.attributes['config']
+ 'poi_name' : self.attributes['poi_names'],
+ 'data_name' : self.attributes['data_name'],
+ 'config' : self.attributes['config'],
+ 'verbosity' : self.attributes['verbosity']
}
self.set_param_ind_kwargs(**param_ind_kwargs)
@@ -234,5 +235,4 @@ class ParameterisedLikelihood(ParameterisedRunner):
outdir = self.attributes['outdir']
outname = self.attributes['outname'].format(poi_names="_".join(poi_names))
outpath = os.path.join(outdir, outname.format(poi_name=poi_name))
- with open(outpath, 'w') as outfile:
- json.dump(data, outfile, indent=3)
\ No newline at end of file
+ save_as_json(data, outpath)
\ No newline at end of file
diff --git a/quickstats/concurrent/parameterised_runner.py b/quickstats/concurrent/parameterised_runner.py
index 824bb4408dda9fc56d1f41fe1c1d9a2a36bc1e7c..531ca677d55be85cb2b25bbf5301d4072003b870 100644
--- a/quickstats/concurrent/parameterised_runner.py
+++ b/quickstats/concurrent/parameterised_runner.py
@@ -1,9 +1,12 @@
from typing import Optional, List, Dict, Union, Tuple
import os
+import ROOT
+
+from quickstats import semistaticmethod
from . import AbstractRunner
from quickstats.parsers import ParamParser
-from quickstats.utils.common_utils import combine_dict
+from quickstats.utils.common_utils import combine_dict, batch_makedirs
class ParameterisedRunner(AbstractRunner):
@@ -19,28 +22,64 @@ class ParameterisedRunner(AbstractRunner):
param_expr:Optional[str]=None,
filter_expr:Optional[str]=None,
exclude_expr:Optional[str]=None,
- parallel:int=-1, timed:bool=True,
+ outdir:str="output", cachedir:str="cache",
save_log:bool=True, cache:bool=True,
- allow_none:bool=True,
+ allow_none:bool=True, parallel:int=-1,
verbosity:Optional[Union[int, str]]="INFO"):
- super().__init__(parallel=parallel, timed=timed, save_log=save_log,
+ super().__init__(parallel=parallel, save_log=save_log,
cache=cache, verbosity=verbosity)
+
+ self.attributes = {
+ 'outdir' : outdir,
+ 'cachedir' : cachedir,
+ 'verbosity': verbosity
+ }
+
self.filter_expr = filter_expr
self.exclude_expr = exclude_expr
self._parser = ParamParser(file_expr, param_expr, allow_none=allow_none)
self._param_points = None
self.param_ind_kwargs = {}
self.param_dpd_kwargs = {}
+
+ def _prerun_batch(self):
+ outdir = self.attributes['outdir']
+ cache_dir = self.get_cache_dir()
+ batch_makedirs([outdir, cache_dir])
+
+ @semistaticmethod
+ def _cached_return(self, outname:str):
+ with open(outname, 'r') as f:
+ result = json.load(f)
+ processed_result = self._process_result(result)
+ return processed_result
+
+ @semistaticmethod
+ def _process_result(self, result:Dict):
+ return result
def setup_parser(self, file_expr:Optional[str]=None,
param_expr:Optional[str]=None):
self._parser.setup(file_expr, param_expr)
+
+ def get_cache_dir(self):
+ return os.path.join(self.attributes['outdir'], self.attributes['cachedir'])
def get_param_points(self, input_path:str):
param_points = self.parser.get_param_points(input_path, filter_expr=self.filter_expr,
exclude_expr=self.exclude_expr)
return param_points
+
+ def get_internal_param_points(self):
+ param_points = self.parser.get_internal_param_points(filter_expr=self.filter_expr,
+ exclude_expr=self.exclude_expr)
+ return param_points
+
+ def get_external_param_points(self, input_path:str):
+ param_points = self.parser.get_external_param_points(input_path, filter_expr=self.filter_expr,
+ exclude_expr=self.exclude_expr)
+ return param_points
def get_serialised_param_data(self, param_points:str, outdir:str="./", outname:str="{param_str}.json"):
filenames = []
@@ -58,10 +97,9 @@ class ParameterisedRunner(AbstractRunner):
parameter_list.append(parameters)
outnames.append(outname)
if len(outnames) != len(set(outnames)):
- raise RuntimeError("output names are not distinct, please check your input. "
- "(this can be due to 1. multiple tasks have the same set of "
- "parameters or 2. multiple zero-parameter tasks with input files of the same"
- " basename)")
+ raise RuntimeError("output names are not distinct, please check your input. (this can be "
+ "due to 1. multiple tasks have the same set of parameters or 2. "
+ "multiple zero-parameter tasks with input files of the same basename)")
serialised_param_data = {
'filenames': filenames,
'outnames': outnames,
@@ -89,7 +127,20 @@ class ParameterisedRunner(AbstractRunner):
def run(self):
kwarg_set, auxiliary_args = self.prepare_task_inputs()
return self.run_batch(kwarg_set, auxiliary_args=auxiliary_args)
-
+
def _end_of_instance_cleanup(self):
- import ROOT
- ROOT.gROOT.CloseFiles()
\ No newline at end of file
+ ROOT.gROOT.CloseFiles()
+
+ @staticmethod
+ def join_param_setup(base_setup:Optional[str]=None, new_setup:Optional[str]=None):
+ components = []
+ for setup in [base_setup, new_setup]:
+ if not setup:
+ continue
+ if isinstance(setup, dict):
+ setup = ParamParser.val_encode_parameters(setup)
+ assert isinstance(setup, str)
+ components.append(setup)
+ if not components:
+ return None
+ return ",".join(components)
\ No newline at end of file
diff --git a/quickstats/concurrent/parameterised_significance.py b/quickstats/concurrent/parameterised_significance.py
new file mode 100644
index 0000000000000000000000000000000000000000..c2b993a7e7cbc5598656f5b589436cd1b2c3ef03
--- /dev/null
+++ b/quickstats/concurrent/parameterised_significance.py
@@ -0,0 +1,167 @@
+import os
+import sys
+import json
+from typing import Optional, Union, Dict, List, Any
+
+from quickstats import semistaticmethod
+from quickstats.parsers import ParamParser
+from quickstats.concurrent import ParameterisedRunner
+from quickstats.utils.common_utils import batch_makedirs, list_of_dict_to_dict_of_list, save_as_json, combine_dict
+from quickstats.maths.numerics import pretty_value
+from quickstats.components import AnalysisBase, AsimovType, AsimovGenerator
+
+class ParameterisedSignificance(ParameterisedRunner):
+
+ def __init__(self, input_path:str,
+ file_expr:Optional[str]=None,
+ param_expr:Optional[str]=None,
+ poi_name:Optional[str]=None,
+ filter_expr:Optional[str]=None,
+ exclude_expr:Optional[str]=None,
+ data_name:str="combData",
+ snapshot_name:Optional[str]=None,
+ mu_exp:float=0.,
+ asimov_type:Optional[int]=None,
+ config:Optional[Dict]=None,
+ outdir:str="output", cachedir:str="cache",
+ outname:str="{param_names}.json", cache:bool=True,
+ save_log:bool=True, parallel:int=-1,
+ verbosity:Optional[Union[int, str]]="INFO"):
+
+ super().__init__(file_expr=file_expr,
+ param_expr=param_expr,
+ filter_expr=filter_expr,
+ exclude_expr=exclude_expr,
+ outdir=outdir, cachedir=cachedir,
+ parallel=parallel, save_log=save_log,
+ cache=cache, verbosity=verbosity)
+
+ self.attributes.update({
+ 'input_path': input_path,
+ 'poi_name': poi_name,
+ 'data_name': data_name,
+ 'snapshot_name': snapshot_name,
+ 'mu_exp': mu_exp,
+ 'asimov_type': asimov_type,
+ 'config': config,
+ 'outname': outname
+ })
+
+ def _prerun_batch(self):
+ self.stdout.tips("When running likelihood fit on an Asimov dataset, remember to restore the global "
+ "observables to their Asimov values by loading the appropriate snapshot. "
+ "Asimov dataset generated on the fly (via the --asimov_type option) will automatically "
+ "load the internally saved snapshot so no user intervention is needed.")
+ super()._prerun_batch()
+
+ @semistaticmethod
+ def _prerun_instance(self, filename:str, parameters:Optional[Union[float, Dict[str, float]]]=None, **kwargs):
+ param_str = "("+ParamParser.val_encode_parameters(parameters)+")"
+ self.stdout.info(f"Evaluating significance {param_str} for the workspace {filename}")
+
+ @semistaticmethod
+ def _run_instance(self, filename:str,
+ poi_name:Optional[str]=None,
+ data_name:str="combData",
+ config:Optional[Dict]=None,
+ mu_exp:float=0.,
+ asimov_type:Optional[int]=None,
+ snapshot_name:Optional[str]=None,
+ outname:Optional[str]=None,
+ **kwargs):
+ try:
+ if config is None:
+ config = {}
+ config.update(kwargs)
+ config['filename'] = filename
+ config['poi_name'] = poi_name
+ config['data_name'] = data_name
+ config['snapshot_name'] = snapshot_name
+ analysis = AnalysisBase(**config)
+ if asimov_type is not None:
+ asimov_type = AsimovType.parse(asimov_type)
+ asimov_data = analysis.generate_standard_asimov(asimov_type, do_import=False)
+ asimov_snapshot = AsimovGenerator.ASIMOV_SETTINGS[asimov_type]['asimov_snapshot']
+ analysis.set_data(asimov_data)
+ result = analysis.nll_fit(poi_val=mu_exp, mode='hybrid',
+ snapshot_name=asimov_snapshot)
+ else:
+ result = analysis.nll_fit(poi_val=mu_exp, mode='hybrid')
+ if outname:
+ with open(outname, 'w') as outfile:
+ json.dump(result, outfile, indent=2)
+ return result
+ except Exception as e:
+ sys.stdout.write(f"{e}\n")
+ return None
+
+ def prepare_task_inputs(self):
+ input_path = self.attributes['input_path']
+ cache_dir = self.get_cache_dir()
+ outname = "{param_str}.json"
+ param_points = self.get_param_points(input_path)
+ param_data = self.get_serialised_param_data(param_points, outdir=cache_dir, outname=outname)
+
+ configs = []
+ base_config = self.attributes['config']
+ if base_config is None:
+ base_config = {}
+ for param_point in param_points:
+ base_fix_params = base_config.get("fix_param", None)
+ internal_params = param_point['internal_parameters']
+ fix_param_expr = self.join_param_setup(base_fix_params, internal_params)
+ config = combine_dict(base_config, {"fix_param": fix_param_expr})
+ configs.append(config)
+ param_dpd_kwargs = {
+ 'parameters' : param_data['parameters'], # just for display
+ 'filename' : param_data['filenames'],
+ 'outname' : param_data['outnames'],
+ 'config' : configs
+ }
+
+ param_ind_kwargs = {}
+ for param in ['poi_name', 'data_name', 'snapshot_name',
+ 'mu_exp', 'asimov_type']:
+ param_ind_kwargs[param] = self.attributes[param]
+
+ self.set_param_ind_kwargs(**param_ind_kwargs)
+ self.set_param_dpd_kwargs(**param_dpd_kwargs)
+ kwarg_set = self.create_kwarg_set()
+ auxiliary_args = {
+ 'parameters': param_data['parameters']
+ }
+ if not kwarg_set:
+ raise RuntimeError("no parameter point to scan for")
+ return kwarg_set, auxiliary_args
+
+ def postprocess(self, raw_result, auxiliary_args:Dict):
+ parameters = auxiliary_args['parameters']
+ data = list_of_dict_to_dict_of_list(parameters)
+ param_names = list(data.keys())
+ data.update({
+ 'nll_muexp' : [],
+ 'nll_muhat' : [],
+ 'qmu' : [],
+ 'muexp' : [],
+ 'muhat' : [],
+ 'significance' : [],
+ 'pvalue' : [],
+ 'status_muexp' : [],
+ 'status_muhat' : [],
+ 'status' : []
+ })
+ for result in raw_result:
+ data['nll_muexp'].append(result['cond_fit']['nll'])
+ data['nll_muhat'].append(result['uncond_fit']['nll'])
+ data['qmu'].append(result['qmu'])
+ data['muexp'].append(next(iter(result['cond_fit']['mu'].values())))
+ data['muhat'].append(next(iter(result['uncond_fit']['muhat'].values())))
+ data['significance'].append(result['significance'])
+ data['pvalue'].append(result['pvalue'])
+ data['status_muexp'].append(result['cond_fit']['status'])
+ data['status_muhat'].append(result['uncond_fit']['status'])
+ data['status'].append(result['status'])
+ outdir = self.attributes['outdir']
+ outname = self.attributes['outname'].format(param_names="_".join(param_names))
+ outpath = os.path.join(outdir, outname)
+ save_as_json(data, outpath)
\ No newline at end of file
diff --git a/quickstats/core/io.py b/quickstats/core/io.py
index 0978f7a8b5a999abd882a483fc1e06e58e567f16..f5e9eeb881200f434bb2423a395697dc7ef89038 100644
--- a/quickstats/core/io.py
+++ b/quickstats/core/io.py
@@ -29,7 +29,8 @@ text_color_map = {
'bright cyan': '\033[36;1m',
'bright white': '\033[37;1m',
'darkred': '\033[91m',
- 'reset': '\033[0m'
+ 'reset': '\033[0m',
+ 'okgreen': '\033[92m'
}
def get_colored_text(text: str, color: str) -> str:
diff --git a/quickstats/core/path_manager.py b/quickstats/core/path_manager.py
index 882060c55809300571551fa10198fd0ce6ba5c24..b300eeb11513d4660c92a3d94e8b22569af9f3f6 100644
--- a/quickstats/core/path_manager.py
+++ b/quickstats/core/path_manager.py
@@ -100,13 +100,17 @@ class PathManager:
def set_file(self, file_name:str, file:Union[str, DynamicFilePath, Tuple[Optional[str], str]]):
self.update_files({file_name: file})
+
+ def get_base_path(self):
+ return self.base_path
- def get_directory(self, directory_name:str, check_exist:bool=False):
+ def get_directory(self, directory_name:str, check_exist:bool=False, **parameters):
if directory_name not in self.directories:
raise KeyError(f'unrecognized directory name "{directory_name}"')
- if self.base_path is not None:
- return os.path.join(self.base_path, self.directories[directory_name])
- directory = self.directories[directory_name]
+ base_path = self.get_base_path()
+ if base_path is not None:
+ return os.path.join(base_path, self.directories[directory_name])
+ directory = self.directories[directory_name].format(**parameters)
if check_exist:
if not os.path.exists(directory):
raise FileNotFoundError(f'directory "{directory}" does not exist')
@@ -147,23 +151,23 @@ class PathManager:
if os.path.isdir(file):
raise ValueError(f'"{file}" is a directory')
- def get_file(self, file_name:str, check_exist:bool=False):
- return self.get_resolved_file(file_name, check_exist=check_exist)
+ def get_file(self, file_name:str, check_exist:bool=False, **parameters):
+ return self.get_resolved_file(file_name, check_exist=check_exist, **parameters)
- def get_directories(self, directory_names:Optional[List[str]]=None):
+ def get_directories(self, directory_names:Optional[List[str]]=None, **parameters):
directory_paths = {}
if directory_names is None:
directory_names = list(self.directories.keys())
for directory_name in directory_names:
- directory_paths[directory_name] = self.get_directory(directory_name)
+ directory_paths[directory_name] = self.get_directory(directory_name, **parameters)
return directory_paths
- def get_files(self, file_names:Optional[List[str]]=None):
+ def get_files(self, file_names:Optional[List[str]]=None, **parameters):
file_paths = {}
if file_names is None:
file_names = list(self.files.keys())
for file_name in file_names:
- file_paths[file_name] = self.get_file(file_name)
+ file_paths[file_name] = self.get_file(file_name, **parameters)
return file_paths
def get_relpath(self, path:str):
@@ -171,39 +175,44 @@ class PathManager:
return path
return os.path.join(self.base_path, path)
- def directory_exists(self, directory_name:str):
- directory = self.get_directory(directory_name)
+ def directory_exists(self, directory_name:str, **parameters):
+ directory = self.get_directory(directory_name, **parameters)
return os.path.exists(directory)
def file_exists(self, file_name:str, **parameters):
file = self.get_resolved_file(file_name, **parameters)
return os.path.exists(file)
- def check_directory(self, directory_name:str):
- self.get_directory(directory_name, check_exist=True)
+ def check_directory(self, directory_name:str, **parameters):
+ self.get_directory(directory_name, **parameters, check_exist=True)
def check_file(self, file_name:str, **parameters):
self.get_resolved_file(file_name, check_exist=True, **parameters)
def makedirs(self, include_names:Optional[List[str]]=None,
- exclude_names:Optional[List[str]]=None):
+ exclude_names:Optional[List[str]]=None,
+ **parameters):
if include_names is None:
include_names = list(self.directories.keys())
if exclude_names is None:
exclude_names = []
- directory_names = list(set(include_names) - set(exclude_names))
- for directory_name in directory_names:
- if directory_name not in self.directories:
- raise KeyError(f'unrecognized directory name "{directory_name}"')
- dirname = self.get_directory(directory_name)
- if not os.path.exists(dirname):
- os.makedirs(dirname)
-
- def makedir_for_files(self, file_names:Union[str, List[str]]):
+ dirnames = list(set(include_names) - set(exclude_names))
+ resolved_dirnames = []
+ for dirname in dirnames:
+ if dirname not in self.directories:
+ raise KeyError(f'unrecognized directory name "{dirname}"')
+ resolved_dirname = self.get_directory(dirname, **parameters)
+ resolved_dirnames.append(resolved_dirname)
+ from quickstats.utils.common_utils import batch_makedirs
+ batch_makedirs(resolved_dirnames)
+
+ def makedir_for_files(self, file_names:Union[str, List[str]], **parameters):
if isinstance(file_names, str):
file_names = [file_names]
- files = self.get_files(file_names)
+ files = self.get_files(file_names, **parameters)
+ resolved_dirnames = []
for file in files.values():
dirname = os.path.dirname(file)
- if (dirname != "") and not os.path.exists(dirname):
- os.makedirs(dirname)
\ No newline at end of file
+ resolved_dirnames.append(dirname)
+ from quickstats.utils.common_utils import batch_makedirs
+ batch_makedirs(resolved_dirnames)
\ No newline at end of file
diff --git a/quickstats/interface/root/ModelConfig.py b/quickstats/interface/root/ModelConfig.py
new file mode 100644
index 0000000000000000000000000000000000000000..8584c3ec1e8792c3c61b2878f0dc423d109813a6
--- /dev/null
+++ b/quickstats/interface/root/ModelConfig.py
@@ -0,0 +1,83 @@
+from typing import Dict, Union, List, Optional, Tuple
+import os
+
+import numpy as np
+
+from quickstats import semistaticmethod, AbstractObject
+from quickstats import GeneralEnum
+from .RooArgSet import RooArgSet
+
+class ObjectType(GeneralEnum):
+
+ PDF = (0, "pdf", "pdf")
+ OBS = (1, "observables", "observable")
+ POIS = (2, "parameters of interest", "parameter of interest")
+ NUIS = (3, "nuisance parameters", "nuisane parameter")
+ GLOBS = (4, "global observables", "global observable")
+
+ def __new__(cls, value:int, plural_str:str, singular_str:str):
+ obj = object.__new__(cls)
+ obj._value_ = value
+ obj.plural_str = plural_str
+ obj.singular_str = singular_str
+ return obj
+
+class ModelConfig(AbstractObject):
+
+ @semistaticmethod
+ def sanity_check(self, mc:"ROOT.RooStats.ModelConfig"):
+ pass_check = True
+ def check_objects(objects, object_type:ObjectType,
+ require_exist:bool=True,
+ require_class:Optional[str]=None,
+ require_dependence:Optional["ROOT.RooAbsArg"]=None,
+ dep_object_type:ObjectType=ObjectType.PDF):
+ if require_exist and (not objects):
+ self.stdout.error(f'{object_type.plural_str} not defined in model config')
+ return False
+ if require_class is not None:
+ assert objects.InheritsFrom("RooArgSet")
+ invalid_objects = RooArgSet.exclude_by_class(objects, require_class)
+ if invalid_objects:
+ for invalid_object in invalid_objects:
+ classname = invalid_object.ClassName()
+ name = invalid_object.GetName()
+ self.stdout.error(f'{object_type.singular_str} "{name}" is an instance of '
+ f'{classname} but not {require_class}')
+ return False
+ if require_dependence is not None:
+ assert objects.InheritsFrom("RooArgSet")
+ assert require_dependence.InheritsFrom("RooAbsArg")
+ dependent_objects = RooArgSet.select_dependent_parameters(objects, require_dependence)
+ invalid_objects = objects.Clone()
+ invalid_objects.remove(dependent_objects)
+ if invalid_objects:
+ for invalid_object in invalid_objects:
+ classname = invalid_object.ClassName()
+ name = invalid_object.GetName()
+ self.stdout.error(f'{dep_object_type.singular_str} does not depend on '
+ f'{object_type.singular_str} "{name}"')
+ return False
+ return True
+ pdf = mc.GetPdf()
+ # check pdf
+ pass_check &= check_objects(pdf, ObjectType.PDF)
+ # skip subsequent checks if pdf does not exist
+ if not pdf:
+ return False
+ # check observables
+ pass_check &= check_objects(mc.GetObservables(), ObjectType.OBS)
+ # check parameters of interest
+ pass_check &= check_objects(mc.GetParametersOfInterest(), ObjectType.POIS,
+ require_class="RooRealVar", require_dependence=pdf)
+ # check nuisance parameters
+ pass_check &= check_objects(mc.GetNuisanceParameters(), ObjectType.NUIS,
+ require_class="RooRealVar", require_dependence=pdf)
+ # check nuisance parameters
+ pass_check &= check_objects(mc.GetNuisanceParameters(), ObjectType.GLOBS,
+ require_class="RooRealVar", require_dependence=pdf)
+
+ # check factorize pdf (needed?)
+
+ return pass_check
+
\ No newline at end of file
diff --git a/quickstats/interface/root/RooAbsPdf.py b/quickstats/interface/root/RooAbsPdf.py
index c2c562da3bc9d0578c9cbd2228d7e2ff3bfe8ec5..ef4ca363ae6f56d92756112e1c0b4fae7d836802 100644
--- a/quickstats/interface/root/RooAbsPdf.py
+++ b/quickstats/interface/root/RooAbsPdf.py
@@ -7,6 +7,8 @@ import numpy as np
from quickstats import semistaticmethod, AbstractObject
from quickstats.interface.cppyy.vectorize import as_np_array, as_vector, np_type_str_maps
from .TH1 import TH1
+from .TH2 import TH2
+from .TH3 import TH3
from .TArrayData import TArrayData
class RooAbsPdf(AbstractObject):
@@ -204,6 +206,9 @@ class RooAbsPdf(AbstractObject):
if channel_cat not in pdf_observables:
pdf_observables.add(channel_cat)
from quickstats.interface.root import RooDataSet
+ # reset initial values of observables
+ for obs in pdf_observables:
+ obs.setBin(0)
c_dataset_map = RooDataSet.get_dataset_map(dataset_map)
asimov_data = ROOT.RooDataSet(dataset_name, dataset_name,
pdf_observables,
@@ -212,8 +217,50 @@ class RooAbsPdf(AbstractObject):
ROOT.RooFit.WeightVar(weight_var))
return asimov_data
+ # has precision issue since the histograms are always of type "float"
+ @semistaticmethod
+ def _get_histo_values_old(self, pdf:"ROOT.RooAbsPdf",
+ obs_x:"ROOT.RooRealVar",
+ obs_y:Optional["ROOT.RooRealVar"]=None,
+ obs_z:Optional["ROOT.RooRealVar"]=None):
+ import ROOT
+ var_y = ROOT.RooFit.YVar(obs_y) if obs_y is not None else ROOT.RooCmdArg.none()
+ var_z = ROOT.RooFit.ZVar(obs_z) if obs_z is not None else ROOT.RooCmdArg.none()
+ rhist = pdf.createHistogram(f'{uuid.uuid4().hex}', obs_x, var_y, var_z)
+ ndim = rhist.GetDimension()
+ if ndim == 1:
+ result = TH1.GetBinContentArray(rhist).flatten()
+ elif ndim == 2:
+ result = TH2.GetBinContentArray(rhist).flatten()
+ elif ndim == 3:
+ result = TH3.GetBinContentArray(rhist).flatten()
+ else:
+ raise RuntimeError('histogram dimension must be 1, 2 or 3')
+ rhist.Delete()
+ return result
+
+ @semistaticmethod
+ def _get_histo_values(self, pdf:"ROOT.RooAbsPdf", observables:"ROOT.RooRealVar"):
+ rhist = self.create_histogram(f'{uuid.uuid4().hex}', observables, 'double')
+ if pdf.canBeExtended():
+ scale_factor = pdf.expectedEvents(observables)
+ else:
+ scale_factor = 1.0
+ pdf.fillHistogram(rhist, observables, scale_factor, 0, False)
+ ndim = rhist.GetDimension()
+ if ndim == 1:
+ result = TH1.GetBinContentArray(rhist).flatten()
+ elif ndim == 2:
+ result = TH2.GetBinContentArray(rhist).flatten()
+ elif ndim == 3:
+ result = TH3.GetBinContentArray(rhist).flatten()
+ else:
+ raise RuntimeError('histogram dimension must be 1, 2 or 3')
+ rhist.Delete()
+ return result
+
@semistaticmethod
- def get_asimov_dataset(self, pdf:"RooAbsPdf",
+ def get_asimov_dataset(self, pdf:"ROOT.RooAbsPdf",
observables:"ROOT.RooArgSet",
weight_name:Optional[str]="weight",
dataset_name:Optional[str]="asimovData",
@@ -243,9 +290,14 @@ class RooAbsPdf(AbstractObject):
np_ds = {}
for i, obs_name in enumerate(bin_centers):
np_ds[obs_name] = bin_center_combination[:, i]
- ds = RooDataSet.from_numpy(np_ds, pdf_observables)
- pdf_values = ROOT.RFUtils.GetPdfValuesAcrossObsDataset(pdf, ds, True)
- pdf_values = as_np_array(pdf_values)
+ nobs = len(bin_centers)
+ # use histogram method for low dimensions
+ if nobs in [1, 2, 3]:
+ pdf_values = self._get_histo_values(pdf, pdf_observables)
+ else:
+ ds = RooDataSet.from_numpy(np_ds, pdf_observables)
+ pdf_values = ROOT.RFUtils.GetPdfValuesAcrossObsDataset(pdf, ds, True)
+ pdf_values = as_np_array(pdf_values)
# multiply by bin width(s)
num_events = pdf_values * np.prod(bin_width_combination, axis=1)
np_ds[weight_name] = num_events
@@ -320,3 +372,62 @@ class RooAbsPdf(AbstractObject):
pdf_title = pdf.GetTitle()
new_pdf = ROOT.RooProdPdf(pdf_name, pdf_title, base_components)
return new_pdf
+
+ @semistaticmethod
+ def create_histogram(self, name:str, observables:"ROOT.RooArgSet", dtype:str="double"):
+ ndim = len(observables)
+ if (ndim < 1) or (ndim > 3):
+ raise ValueError('dimension not supported')
+ binnings = [observable.getBinning() for observable in observables]
+ if dtype not in ["float", "double"]:
+ raise RuntimeError('dtype must be "float" or "double"')
+ import ROOT
+ if ndim == 1:
+ cls = ROOT.TH1D if dtype == "double" else ROOT.TH1F
+ if binnings[0].isUniform():
+ histogram = cls(name, name,
+ binnings[0].numBins(),
+ binnings[0].lowBound(),
+ binnings[0].highBound())
+ else:
+ histogram = cls(name, name,
+ binnings[0].numBins(),
+ binnings[0].array())
+ elif ndim == 2:
+ cls = ROOT.TH2D if dtype == "double" else ROOT.TH2F
+ if (binnings[0].isUniform() and binnings[1].isUniform()):
+ histogram = cls(name, name,
+ binnings[0].numBins(),
+ binnings[0].lowBound(),
+ binnings[0].highBound(),
+ binnings[1].numBins(),
+ binnings[1].lowBound(),
+ binnings[1].highBound())
+ else:
+ histogram = cls(name, name,
+ binnings[0].numBins(),
+ binnings[0].array(),
+ binnings[1].numBins(),
+ binnings[1].array())
+ else:
+ cls = ROOT.TH3D if dtype == "double" else ROOT.TH3F
+ if (binnings[0].isUniform() and binnings[1].isUniform() and binnings[2].isUniform()):
+ histogram = cls(name, name,
+ binnings[0].numBins(),
+ binnings[0].lowBound(),
+ binnings[0].highBound(),
+ binnings[1].numBins(),
+ binnings[1].lowBound(),
+ binnings[1].highBound(),
+ binnings[2].numBins(),
+ binnings[2].lowBound(),
+ binnings[2].highBound())
+ else:
+ histogram = cls(name, name,
+ binnings[0].numBins(),
+ binnings[0].array(),
+ binnings[1].numBins(),
+ binnings[1].array(),
+ binnings[2].numBins(),
+ binnings[2].array())
+ return histogram
diff --git a/quickstats/interface/root/RooArgSet.py b/quickstats/interface/root/RooArgSet.py
index e29fdf2d058f2ccd0bf70a049feca6db90e14b64..6f86bbc23f71e9566ad60fb358c4e5a0032dec41 100644
--- a/quickstats/interface/root/RooArgSet.py
+++ b/quickstats/interface/root/RooArgSet.py
@@ -2,8 +2,10 @@ from typing import Dict, Union, List, Optional
import numpy as np
+import cppyy
+
from quickstats import semistaticmethod
-from quickstats.interface.cppyy.vectorize import as_np_array
+from quickstats.interface.cppyy.vectorize import as_vector
class RooArgSet:
@@ -15,4 +17,56 @@ class RooArgSet:
@staticmethod
def sort(argset:"ROOT.RooArgSet"):
argset.sort()
- return argset
\ No newline at end of file
+ return argset
+
+ def get_boundary_parameters(argset:"ROOT.RooArgSet"):
+ return cppyy.gbl.RFUtils.GetBoundaryParameters(argset)
+
+ def select_by_class(argset:"ROOT.RooArgSet", classname:str):
+ return cppyy.gbl.RFUtils.SelectByClass(argset, classname)
+
+ def exclude_by_class(argset:"ROOT.RooArgSet", classname:str):
+ return cppyy.gbl.RFUtils.ExcludeByClass(argset, classname)
+
+ def select_dependent_parameters(argset:"ROOT.RooArgSet", source:"ROOT.RooAbsArg"):
+ return cppyy.gbl.RFUtils.SelectDependentParameters(argset, source)
+
+ def get_set_difference(argset1:"ROOT.RooArgSet", argset2:"ROOT.RooArgSet"):
+ return cppyy.gbl.RFUtils.GetRooArgSetDifference(argset1, argset2)
+
+ def set_constant_state(argset:"ROOT.RooArgSet", value:bool=True):
+ argset.setAttribAll('Constant', value)
+
+ def select_by_constant_state(argset:"ROOT.RooArgSet", value:bool=True):
+ return argset.selectByAttrib('Constant', value)
+
+ def get_parameters_close_to_min(argset:"ROOT.RooArgSet", threshold:float=0.1):
+ return cppyy.gbl.RFUtils.GetParametersCloseToMin(argset, threshold)
+
+ def get_parameters_close_to_max(argset:"ROOT.RooArgSet", threshold:float=0.1):
+ return cppyy.gbl.RFUtils.GetParametersCloseToMax(argset, threshold)
+
+ def get_parameters_close_to_boundary(argset:"ROOT.RooArgSet", threshold:float=0.1):
+ return cppyy.gbl.RFUtils.GetParametersCloseToBoundary(argset, threshold)
+
+ def expand_parameters_range(argset:"ROOT.RooArgSet", threshold:float=0.1,
+ expand_min:bool=True, expand_max:bool=True,
+ orig_argset_at_min:Optional["ROOT.RooArgSet"]=None,
+ new_argset_at_min:Optional["ROOT.RooArgSet"]=None,
+ orig_argset_at_max:Optional["ROOT.RooArgSet"]=None,
+ new_argset_at_max:Optional["ROOT.RooArgSet"]=None,):
+ orig_argset_at_min = 0 if orig_argset_at_min is None else orig_argset_at_min
+ new_argset_at_min = 0 if new_argset_at_min is None else new_argset_at_min
+ orig_argset_at_max = 0 if orig_argset_at_max is None else orig_argset_at_max
+ new_argset_at_max = 0 if new_argset_at_max is None else new_argset_at_max
+ return cppyy.gbl.RFUtils.ExpandParametersRange(argset, threshold,
+ expand_min, expand_max,
+ orig_argset_at_min,
+ new_argset_at_min,
+ orig_argset_at_max,
+ new_argset_at_max)
+ def set_category_indices(argset:"ROOT.RooArgSet", indices:np.ndarray):
+ return cppyy.gbl.RFUtils.SetCategoryIndices(argset, as_vector(indices))
+
+ def save_data_as_txt(argset:"ROOT.RooArgSet", filename:str, precision:int=7):
+ return cppyy.gbl.RFUtils.SaveRooArgSetDataAsTxt(argset, filename, precision)
\ No newline at end of file
diff --git a/quickstats/interface/root/RooRealVar.py b/quickstats/interface/root/RooRealVar.py
index 672c5dc4a1825566d9a749ba81c2649440ad8e9c..daae909dd3f155c6cc71060311ca45e0be7870ad 100644
--- a/quickstats/interface/root/RooRealVar.py
+++ b/quickstats/interface/root/RooRealVar.py
@@ -5,6 +5,8 @@ from quickstats import semistaticmethod
from quickstats.maths.numerics import get_proper_ranges
from quickstats.interface.root import TArrayData
+import cppyy
+
class RooRealVar:
@property
@@ -158,4 +160,20 @@ class RooRealVar:
'bin_center': bin_edge_to_bin_center(bin_low_edge),
'bin_width': bin_edge_to_bin_width(bin_low_edge)
}
- return result
\ No newline at end of file
+ return result
+
+ @staticmethod
+ def at_boundary(obj:"ROOT.RooRealVar"):
+ return cppyy.gbl.RFUtils.ParameterAtBoundary(obj)
+
+ @staticmethod
+ def close_to_min(obj:"ROOT.RooRealVar", threshold:float=0.1):
+ return cppyy.gbl.RFUtils.ParameterCloseToMin(obj, threshold)
+
+ @staticmethod
+ def close_to_min(obj:"ROOT.RooRealVar", threshold:float=0.1):
+ return cppyy.gbl.RFUtils.ParameterCloseToMax(obj, threshold)
+
+ @staticmethod
+ def close_to_boundary(obj:"ROOT.RooRealVar", threshold:float=0.1):
+ return cppyy.gbl.RFUtils.ParameterCloseToBoundary(obj, threshold)
\ No newline at end of file
diff --git a/quickstats/interface/root/TH2.py b/quickstats/interface/root/TH2.py
index e74e71e8945090d80f7671333e75520f60dc3da6..391e4f5fd832c802c31a526daf00cea3bfcdfa3b 100644
--- a/quickstats/interface/root/TH2.py
+++ b/quickstats/interface/root/TH2.py
@@ -27,12 +27,12 @@ class TH2(TObject):
def get_fundamental_type(self):
import ROOT
- raise ROOT.TH2
+ return ROOT.TH2
def init(self, h):
self.bin_content = self.GetBinContentArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
- self.x_labels = self.GetXLabelArray(h)
- self.y_labels = self.GetYLabelArray(h)
+ #self.x_labels = self.GetXLabelArray(h)
+ #self.y_labels = self.GetYLabelArray(h)
self.x_bin_center = self.GetXBinCenterArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
self.y_bin_center = self.GetYBinCenterArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
self.x_bin_width = self.GetXBinWidthArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
@@ -47,7 +47,8 @@ class TH2(TObject):
n_bins_y = h.GetNbinsY()
# account for underflow and overflow bins
size = (n_bins_x + 2) * (n_bins_y + 2)
- np_arr = c_array_to_np_array(arr, size=size, shape=(n_bins_x + 2, n_bins_y + 2))
+ np_arr = c_array_to_np_array(arr, size=size)
+ np_arr = np_arr.reshape((n_bins_x + 2, n_bins_y + 2), order='F')
x_start = 1 if not underflow_bin else 0
y_start = x_start
x_end = -1 if not overflow_bin else n_bins_x + 2
diff --git a/quickstats/interface/root/TH3.py b/quickstats/interface/root/TH3.py
new file mode 100644
index 0000000000000000000000000000000000000000..fc7d9ad0d8cd2e51ea091cdff2ffb9893738325f
--- /dev/null
+++ b/quickstats/interface/root/TH3.py
@@ -0,0 +1,127 @@
+import array
+import numpy as np
+
+from quickstats import semistaticmethod
+from quickstats.interface.root import TObject, TArrayData
+from quickstats.interface.cppyy.vectorize import c_array_to_np_array
+
+class TH3(TObject):
+
+ DTYPE_MAP = {
+ "TH3I": "int",
+ "TH3F": "float",
+ "TH3D": "double"
+ }
+
+ def __init__(self, h:"ROOT.TH3", underflow_bin:bool=False, overflow_bin:bool=False):
+ import ROOT
+ dtype_map = {
+ ROOT.TH3I: "int",
+ ROOT.TH3F: "float",
+ ROOT.TH3D: "double"
+ }
+ self.dtype = dtype_map.get(type(h), "double")
+ self.underflow_bin = underflow_bin
+ self.overflow_bin = overflow_bin
+ self.init(h)
+
+ def get_fundamental_type(self):
+ import ROOT
+ return ROOT.TH3
+
+ def init(self, h):
+ self.bin_content = self.GetBinContentArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
+ self.x_bin_center = self.GetXBinCenterArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
+ self.y_bin_center = self.GetYBinCenterArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
+ self.z_bin_center = self.GetZBinCenterArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
+ self.x_bin_width = self.GetXBinWidthArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
+ self.y_bin_width = self.GetYBinWidthArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
+ self.z_bin_width = self.GetZBinWidthArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
+ self.x_bin_low_edge = self.GetXBinLowEdgeArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
+ self.y_bin_low_edge = self.GetYBinLowEdgeArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
+ self.z_bin_low_edge = self.GetZBinLowEdgeArray(h, self.dtype, self.underflow_bin, self.overflow_bin)
+
+ @staticmethod
+ def GetBinContentArray(h, dtype:str='double', underflow_bin:bool=False, overflow_bin:bool=False):
+ arr = h.GetArray()
+ n_bins_x = h.GetNbinsX()
+ n_bins_y = h.GetNbinsY()
+ n_bins_z = h.GetNbinsZ()
+ # account for underflow and overflow bins
+ size = (n_bins_x + 2) * (n_bins_y + 2) * (n_bins_z + 2)
+ np_arr = c_array_to_np_array(arr, size=size)
+ np_arr = np_arr.reshape((n_bins_x + 2, n_bins_y + 2, n_bins_z + 2), order='F')
+ x_start = 1 if not underflow_bin else 0
+ y_start = x_start
+ z_start = x_start
+ x_end = -1 if not overflow_bin else n_bins_x + 2
+ y_end = -1 if not overflow_bin else n_bins_y + 2
+ z_end = -1 if not overflow_bin else n_bins_z + 2
+ np_arr = np_arr[x_start:x_end, y_start:y_end, z_start:z_end]
+ return np_arr
+
+ @staticmethod
+ def GetXLabelArray(h:"ROOT.TH2"):
+ return np.array(h.GetXaxis().GetLabels(), dtype=str)
+
+ @staticmethod
+ def GetYLabelArray(h:"ROOT.TH2"):
+ return np.array(h.GetYaxis().GetLabels(), dtype=str)
+
+ @staticmethod
+ def GetZLabelArray(h:"ROOT.TH2"):
+ return np.array(h.GetYaxis().GetLabels(), dtype=str)
+
+ @staticmethod
+ def GetAxisBinCenterArray(ax:"ROOT.TAxis", dtype:str='double', underflow_bin:int=0, overflow_bin:int=0):
+ import ROOT
+ c_vector = ROOT.TAxisUtils.GetBinCenterArray[dtype](ax, underflow_bin, overflow_bin)
+ return TArrayData.vec_to_array(c_vector)
+
+ @staticmethod
+ def GetAxisBinWidthArray(ax:"ROOT.TAxis", dtype:str='double', underflow_bin:int=0, overflow_bin:int=0):
+ import ROOT
+ c_vector = ROOT.TAxisUtils.GetBinWidthArray[dtype](ax, underflow_bin, overflow_bin)
+ return TArrayData.vec_to_array(c_vector)
+
+ @staticmethod
+ def GetAxisBinLowEdgeArray(ax:"ROOT.TAxis", dtype:str='double', underflow_bin:int=0, overflow_bin:int=0):
+ import ROOT
+ c_vector = ROOT.TAxisUtils.GetBinLowEdgeArray[dtype](ax, underflow_bin, overflow_bin)
+ return TArrayData.vec_to_array(c_vector)
+
+ @staticmethod
+ def GetXBinCenterArray(h:"ROOT.TH2", dtype:str='double', underflow_bin:bool=False, overflow_bin:bool=False):
+ return TH2.GetAxisBinCenterArray(h.GetXaxis(), dtype, underflow_bin, overflow_bin)
+
+ @staticmethod
+ def GetYBinCenterArray(h:"ROOT.TH2", dtype:str='double', underflow_bin:bool=False, overflow_bin:bool=False):
+ return TH2.GetAxisBinCenterArray(h.GetYaxis(), dtype, underflow_bin, overflow_bin)
+
+ @staticmethod
+ def GetZBinCenterArray(h:"ROOT.TH2", dtype:str='double', underflow_bin:bool=False, overflow_bin:bool=False):
+ return TH2.GetAxisBinCenterArray(h.GetZaxis(), dtype, underflow_bin, overflow_bin)
+
+ @staticmethod
+ def GetXBinWidthArray(h:"ROOT.TH2", dtype:str='double', underflow_bin:bool=False, overflow_bin:bool=False):
+ return TH2.GetAxisBinWidthArray(h.GetXaxis(), dtype, underflow_bin, overflow_bin)
+
+ @staticmethod
+ def GetYBinWidthArray(h:"ROOT.TH2", dtype:str='double', underflow_bin:bool=False, overflow_bin:bool=False):
+ return TH2.GetAxisBinWidthArray(h.GetYaxis(), dtype, underflow_bin, overflow_bin)
+
+ @staticmethod
+ def GetZBinWidthArray(h:"ROOT.TH2", dtype:str='double', underflow_bin:bool=False, overflow_bin:bool=False):
+ return TH2.GetAxisBinWidthArray(h.GetZaxis(), dtype, underflow_bin, overflow_bin)
+
+ @staticmethod
+ def GetXBinLowEdgeArray(h:"ROOT.TH2", dtype:str='double', underflow_bin:int=0, overflow_bin:int=0):
+ return TH2.GetAxisBinLowEdgeArray(h.GetXaxis(), dtype, underflow_bin, overflow_bin)
+
+ @staticmethod
+ def GetYBinLowEdgeArray(h:"ROOT.TH2", dtype:str='double', underflow_bin:int=0, overflow_bin:int=0):
+ return TH2.GetAxisBinLowEdgeArray(h.GetYaxis(), dtype, underflow_bin, overflow_bin)
+
+ @staticmethod
+ def GetZBinLowEdgeArray(h:"ROOT.TH2", dtype:str='double', underflow_bin:int=0, overflow_bin:int=0):
+ return TH2.GetAxisBinLowEdgeArray(h.GetZaxis(), dtype, underflow_bin, overflow_bin)
\ No newline at end of file
diff --git a/quickstats/interface/root/__init__.py b/quickstats/interface/root/__init__.py
index b1c921e16e6590ca5d2b8501c8eef60b809cacab..413ffcd9398f76edfab0ccfd80bc0200e55ff657 100644
--- a/quickstats/interface/root/__init__.py
+++ b/quickstats/interface/root/__init__.py
@@ -15,6 +15,7 @@ from quickstats.interface.root.RooCategory import RooCategory
from quickstats.interface.root.RooAbsPdf import RooAbsPdf
from quickstats.interface.root.RooArgSet import RooArgSet
from quickstats.interface.root.RooMsgService import RooMsgService
+from quickstats.interface.root.ModelConfig import ModelConfig
load_macros()
diff --git a/quickstats/interface/root/macros.py b/quickstats/interface/root/macros.py
index 1a84304b64afb1d362d5e3500452d2f4eb5eff80..999b3b76ae270e84ff1e2bb7a714ad5757902f92 100644
--- a/quickstats/interface/root/macros.py
+++ b/quickstats/interface/root/macros.py
@@ -197,10 +197,11 @@ ROOT_MACROS = \
"RFUtils":
"""
#include <iostream>
- #include <chrono>
- using std::chrono::high_resolution_clock;
- using std::chrono::duration;
- using std::chrono::milliseconds;
+ #include <fstream>
+ //#include <chrono>
+ //using std::chrono::high_resolution_clock;
+ //using std::chrono::duration;
+ //using std::chrono::milliseconds;
namespace RFUtils{
struct DatasetStruct {
@@ -413,16 +414,22 @@ ROOT_MACROS = \
const size_t numEntries = dataset->numEntries();
std::vector<double> values(numEntries);
RooArgSet* pdfObs = pdf->getObservables(dataset->get());
- double expected_events = 1.;
+ for (auto const& obs: *pdfObs)
+ obs->recursiveRedirectServers(*pdfObs, false, false, false);
+ RooArgSet projectedVars;
+ RooArgSet* cloneSet = nullptr;
+ const RooAbsReal *projected = pdf->createPlotProjection(*pdfObs, projectedVars, cloneSet);
+ double scale_factor = 1.;
if (normalize)
- expected_events = pdf->expectedEvents(*pdfObs);
+ scale_factor = pdf->expectedEvents(*pdfObs);
for (size_t i = 0; i < numEntries; i++){
// faster than
// pdfObs->assignFast(*dataset->get(i));
for (auto const& x : *dataset->get(i))
((RooRealVar*)(pdfObs->find(x->GetName())))->setVal(((RooRealVar*)x)->getVal());
- values[i] = pdf->getVal(*pdfObs) * expected_events;
+ values[i] = projected->getVal() * scale_factor;
}
+ delete cloneSet;
return values;
}
@@ -460,15 +467,163 @@ ROOT_MACROS = \
}
}
- RooArgSet SelectConstantArgs(RooArgSet* args, bool isConstant=true){
+ RooArgSet GetConstantParameters(RooArgSet* args, bool isConstant=true){
RooArgSet result;
- const size_t n_args = args->size();
- for (size_t i = 0; i < n_args; i++)
- if (((RooRealVar*)(*args)[i])->isConstant() == isConstant)
- result.add(*((*args)[i]));
+ for (auto const *arg: *args)
+ if (((RooRealVar*)arg)->isConstant() == isConstant)
+ result.add(*arg);
return result;
}
+ bool ParameterCloseToMin(RooRealVar* param, float threshold=0.1){
+ if ((!param->hasMin()) || (!param->hasMax()))
+ return false;
+ return param->getVal() < ((1. - threshold) * param->getMin() + threshold * param->getMax());
+ }
+
+ bool ParameterCloseToMax(RooRealVar* param, float threshold=0.1){
+ if ((!param->hasMin()) || (!param->hasMax()))
+ return false;
+ return param->getVal() > ((1. - threshold) * param->getMax() + threshold * param->getMin());
+ }
+
+ bool ParameterCloseToBoundary(RooRealVar* param, float threshold=0.1){
+ return ParameterCloseToMin(param, threshold) || ParameterCloseToMax(param, threshold);
+ }
+
+ RooArgSet GetParametersCloseToMin(RooArgSet* args, float threshold=0.1){
+ RooArgSet result;
+ for (auto const *arg: *args)
+ if (ParameterCloseToMin((RooRealVar*)arg), threshold)
+ result.add(*arg);
+ return result;
+ }
+
+ RooArgSet GetParametersCloseToMax(RooArgSet* args, float threshold=0.1){
+ RooArgSet result;
+ for (auto const *arg: *args)
+ if (ParameterCloseToMax((RooRealVar*)arg), threshold)
+ result.add(*arg);
+ return result;
+ }
+
+ RooArgSet GetParametersCloseToBoundary(RooArgSet* args, float threshold=0.1){
+ RooArgSet result;
+ for (auto const *arg: *args)
+ if (ParameterCloseToBoundary((RooRealVar*)arg), threshold)
+ result.add(*arg);
+ return result;
+ }
+
+ void ExpandParametersRange(RooArgSet* args, float threshold=0.1,
+ bool expand_min=true, bool expand_max=true,
+ RooArgSet* outOrigArgsAtMin=nullptr,
+ RooArgSet* outNewArgsAtMin=nullptr,
+ RooArgSet* outOrigArgsAtMax=nullptr,
+ RooArgSet* outNewArgsAtMax=nullptr){
+ for (auto *arg: *args){
+ RooRealVar* param = dynamic_cast<RooRealVar*>(arg);
+ if (!param)
+ continue;
+ if (expand_min && ParameterCloseToMin(param, threshold)){
+ const double val = param->getVal();
+ if (outOrigArgsAtMin != nullptr)
+ outOrigArgsAtMin->add(*(RooRealVar*)param->Clone());
+ param->setMin(val - (param->getMax() - val));
+ if (outNewArgsAtMin != nullptr)
+ outNewArgsAtMin->add(*(RooRealVar*)param->Clone());
+ }
+ else if (expand_max && ParameterCloseToMax(param, threshold)){
+ const double val = param->getVal();
+ if (outOrigArgsAtMax != nullptr)
+ outOrigArgsAtMax->add(*(RooRealVar*)param->Clone());
+ param->setMax(val + (val - param->getMin()));
+ if (outNewArgsAtMax != nullptr)
+ outNewArgsAtMax->add(*(RooRealVar*)param->Clone());
+ }
+ }
+ }
+
+ bool ParameterAtBoundary(RooRealVar* param, float nsigma=1.0){
+ const double value = param->getVal();
+ return ((value - param->getMin()) < nsigma * -1. * param->getErrorLo()) ||
+ ((param->getMax() - value) < nsigma * param->getErrorHi());
+ }
+
+ RooArgSet GetBoundaryParameters(RooArgSet* args){
+ RooArgSet result;
+ for (auto const *arg: *args)
+ if (ParameterAtBoundary((RooRealVar*)arg))
+ result.add(*arg);
+ return result;
+ }
+
+ RooArgSet SelectByClass(RooArgSet* args, const char* classname){
+ RooArgSet result;
+ for (auto const *arg: *args)
+ if (arg->InheritsFrom(classname))
+ result.add(*arg);
+ return result;
+ }
+
+ RooArgSet ExcludeByClass(RooArgSet* args, const char* classname){
+ RooArgSet result;
+ for (auto const *arg: *args)
+ if (!arg->InheritsFrom(classname))
+ result.add(*arg);
+ return result;
+ }
+
+ RooArgSet SelectDependentParameters(RooArgSet* args, RooAbsArg* source){
+ RooArgSet result;
+ for (auto const *arg: *args)
+ if (source->dependsOn(*arg))
+ result.add(*arg);
+ return result;
+ }
+
+ RooArgSet GetRooArgSetDifference(RooArgSet* args1, RooArgSet* args2){
+ RooArgSet result;
+ for (auto const *arg: *args1)
+ if (!args2->find(*arg))
+ result.add(*arg);
+ return result;
+ }
+
+ template<typename T>
+ int SetCategoryIndices(RooArgList* cats, std::vector<T> *indices){
+ int changedIndex = -1;
+ if (cats->size() != indices->size())
+ throw std::runtime_error("category index size mismatch");
+ for (size_t i = 0; i < indices->size(); ++i){
+ RooCategory* cat = dynamic_cast<RooCategory*>(cats->at(i));
+ if (!cat)
+ throw std::runtime_error("encountered non-RooCategory instance");
+ if (cat->getIndex() != indices->at(i)){
+ changedIndex = i;
+ cat->setIndex(indices->at(i));
+ }
+ }
+ return changedIndex;
+ }
+
+ void SaveRooArgSetDataAsTxt(const RooArgSet *args, const std::string &filename, size_t precision=8){
+ std::ofstream outfile(filename, std::ofstream::trunc);
+ for (auto arg: *args){
+ RooRealVar* v = dynamic_cast<RooRealVar*>(arg);
+ if (v)
+ outfile << v->GetName() <<" "<<std::fixed<<std::setprecision(precision)<<v->getVal()<<" "<<v->isConstant()<<" "<<
+ std::fixed<<std::setprecision(precision)<<v->getMin()<<" "<<
+ std::fixed<<std::setprecision(precision)<<v->getMax()<<std::endl;
+ else{
+ RooCategory* cat = dynamic_cast<RooCategory*>(arg);
+ if (cat)
+ outfile<<cat->GetName()<<" "<<cat->getCurrentIndex()<<" "<<cat->isConstant()<<std::endl;
+ }
+ }
+ outfile.close();
+ }
+
#if ROOT_VERSION_CODE >= ROOT_VERSION(6,26,0)
#endif
};
diff --git a/quickstats/maths/interpolation.py b/quickstats/maths/interpolation.py
index 2a9cbac9a0f7ff2806f0727390c33dffbd29cb89..c3552766b55217143ef66f70d97a0cae703081aa 100644
--- a/quickstats/maths/interpolation.py
+++ b/quickstats/maths/interpolation.py
@@ -54,6 +54,33 @@ def get_x_intersections(x1, y1, x2, y2):
intersections = [x2[i] - (x2[i + 1] - x2[i])/(diff[i + 1] - diff[i]) * diff[i] for i in idx]
return intersections
+def get_intervals_between_curves(x1, y1, x2, y2):
+ """Get x intervals of intersection between two curves
+ """
+ interp_y1 = np.interp(x2, x1, y1)
+
+ diff = interp_y1 - y2
+ sign_change = np.diff(np.sign(diff))
+ # determines what index intersection points are at
+ idx = np.argwhere(sign_change).flatten()
+ #linear interpolation to get exact intercepts: x = x1 + (x2-x1)/(y2-y1) * (y-y1)
+ #y = 0 -> x = x1 - (x2-x1)/(y2-y1) * y1
+ intersections = np.array([x2[i] - (x2[i + 1] - x2[i])/(diff[i + 1] - diff[i]) * diff[i] for i in idx])
+ # no intersection
+ if len(intersections) == 0:
+ return intersections
+ # one-sided interval
+ elif len(intersections) == 1:
+ sign = sign_change[idx[0]]
+ if sign < 0:
+ return np.array([-np.inf, intersections[0]])
+ return np.array([intersections[0], np.inf])
+ elif len(intersections == 2):
+ if (sign_change[idx[0]] + sign_change[idx[1]]) != 0:
+ raise RuntimeError('found discontinuous curves')
+ return intersections
+ raise RuntimeError('found multiple intervals')
+
def interpolate_2d(x:np.ndarray, y:np.ndarray, z:np.ndarray, method:str='cubic', n:int=500):
from scipy import interpolate
mask = ~np.isnan(z)
diff --git a/quickstats/parsers/param_parser.py b/quickstats/parsers/param_parser.py
index f206b8bb67e251de78837e6c262fc428700f2ec2..66311d7788e1baffdface960ac794ff774deb15e 100644
--- a/quickstats/parsers/param_parser.py
+++ b/quickstats/parsers/param_parser.py
@@ -6,6 +6,7 @@ import glob
import fnmatch
import itertools
from functools import partial
+from collections import ChainMap
import numpy as np
@@ -132,7 +133,9 @@ class ParamParser:
return regex
@staticmethod
- def sort_param_points(param_points:List, attributes:List):
+ def sort_param_points(param_points:List, attributes:Optional[List]=None):
+ if attributes is None:
+ attributes = list(param_points[0]['parameters'])
key = lambda d: tuple(d['parameters'][attrib] for attrib in attributes)
return sorted(param_points, key=key)
@@ -176,13 +179,14 @@ class ParamParser:
if param_str is None:
return {}
param_points = []
- idp_param_strs = split_str(param_str, sep=';', remove_empty=True)
+ # components separated by ; are subject to independent conditions
+ ind_components = split_str(param_str, sep=';', remove_empty=True)
all_combinations = []
- for idp_param_str in idp_param_strs:
+ for component in ind_components:
param_values = {}
- param_expr_list = split_str(idp_param_str, sep=',', remove_empty=True)
+ param_expr_list = split_str(component, sep=',', remove_empty=True, use_paranthesis=True)
for expr in param_expr_list:
- tokens = expr.split('=')
+ tokens = split_str(expr, sep='=')
# floating poi
if len(tokens) == 1:
param_name = tokens[0]
@@ -193,26 +197,31 @@ class ParamParser:
continue
if len(tokens) != 2:
raise ValueError('invalid expression for parameterisation')
- param_name = tokens[0]
- values_expr = tokens[1]
- tokens = split_str(values_expr, sep='_')
- # fixed value
- if len(tokens) == 1:
- values = [float(tokens[0])]
- # scan across range
- elif len(tokens) == 3:
- poi_min = float(tokens[0])
- poi_max = float(tokens[1])
- poi_step = float(tokens[2])
- values = np.arange(poi_min, poi_max + poi_step, poi_step)
- else:
- raise ValueError('invalid expression for parameterisation')
+ param_name, values_expr = tokens
if param_name not in param_values:
param_values[param_name] = np.array([])
if None in param_values[param_name]:
raise RuntimeError(f"the parameter {param_name} is being profiled and non-profiled at the same time "
f"in the parameter expression: {idp_param_str}")
- param_values[param_name] = np.concatenate([param_values[param_name], values])
+ # grouped expression
+ if values_expr.startswith('(') and values_expr.endswith(')'):
+ values_expr_list = split_str(values_expr[1:-1], sep=',')
+ else:
+ values_expr_list = [values_expr]
+ for expr in values_expr_list:
+ tokens = split_str(expr, sep='_')
+ # fixed value
+ if len(tokens) == 1:
+ values = [float(tokens[0])]
+ # scan across range
+ elif len(tokens) == 3:
+ poi_min = float(tokens[0])
+ poi_max = float(tokens[1])
+ poi_step = float(tokens[2])
+ values = np.arange(poi_min, poi_max + poi_step, poi_step)
+ else:
+ raise ValueError('invalid expression for parameterisation')
+ param_values[param_name] = np.concatenate([param_values[param_name], values])
param_names = list(param_values.keys())
# convert to numpy arrays
combinations = [np.array(param_values[param_name]) for param_name in param_names]
@@ -274,14 +283,14 @@ class ParamParser:
param_points.append(point)
attributes = list(self.attribute_parser)
param_points = self.sort_param_points(param_points, attributes)
- selected_param_points = self.select_param_points(param_points, filter_expr,
- exclude_expr, dict_key="parameters")
+ selected_param_points = self.get_selected_points(param_points, filter_expr,
+ exclude_expr, dict_keys="parameters")
return selected_param_points
def get_internal_param_points(self, filter_expr:Optional[str]=None,
exclude_expr:Optional[str]=None):
param_points = self.parse_param_str(self.param_str)
- selected_param_points = self.select_param_points(param_points, filter_expr,
+ selected_param_points = self.get_selected_points(param_points, filter_expr,
exclude_expr)
return selected_param_points
@@ -291,66 +300,65 @@ class ParamParser:
return []
expr = remove_whitespace(expr)
match_conditions = []
- tokens = split_str(expr, sep=';', remove_empty=True)
- for token in tokens:
+ # components separated by ; are subject to independent conditions
+ ind_components = split_str(expr, sep=';', remove_empty=True)
+ for component in ind_components:
+ conditions = split_str(component, sep=',', remove_empty=True, use_paranthesis=True)
match_condition = {}
- param_conditions = split_str(token, sep=',', remove_empty=True)
- for param_condition in param_conditions:
- subtokens = split_str(param_condition, sep='=')
- if len(subtokens) != 2:
- raise RuntimeError(f"invalid parameter condition `{param_condition}` in the "
- f"filter/exclude expression `{expr}`")
- param_name, condition = subtokens
- if "*" in condition:
- match_condition[param_name] = partial(MatchMode.Wildcard.match_func, y=condition)
+ for condition in conditions:
+ tokens = split_str(condition, sep='=')
+ if len(tokens) != 2:
+ raise ValueError(f'invalid sub-expression: {condition}')
+ param_name, values_expr = tokens
+ if param_name not in match_condition:
+ match_condition[param_name] = []
+ if values_expr.startswith('(') and values_expr.endswith(')'):
+ values_expr_list = split_str(values_expr[1:-1], sep=',')
else:
- match_condition[param_name] = partial(MatchMode.Numerical.match_func, y=condition)
+ values_expr_list = [values_expr]
+ for values_expr in values_expr_list:
+ if "*" in values_expr:
+ match_func = MatchMode.Wildcard.match_func
+ else:
+ match_func = MatchMode.Numerical.match_func
+ match_condition[param_name].append(partial(match_func, y=values_expr))
match_conditions.append(match_condition)
return match_conditions
-
- @staticmethod
- def select_param_point(parameters, conditions:List[Dict], default:bool=True):
- selected = not default
- for condition in conditions:
- for param_name, match_func in condition.items():
- if param_name not in parameters:
- continue
- param_value = parameters[param_name]
- if not match_func(param_value):
- return not default
- return default
-
- @staticmethod
- def select_param_point(parameters, conditions:List[Dict], default:bool=True):
- for condition in conditions:
- for param_name, match_func in condition.items():
- # consider not matched if any of the param-specific condition not satisfied
- if (param_name not in parameters) or (not match_func(parameters[param_name])):
- break
- else:
- return default
- return not default
-
- def select_param_points(self, param_points:List,
+
+ @semistaticmethod
+ def get_selected_points(self, param_points:List,
filter_expr:Optional[str]=None,
exclude_expr:Optional[str]=None,
- dict_key:Optional=None):
- selected_param_points = []
+ dict_keys:Optional=None):
filter_conditions = self.parse_filter_expr(filter_expr)
exclude_conditions = self.parse_filter_expr(exclude_expr)
- for param_point in param_points:
- if dict_key:
- parameters = param_point[dict_key]
- else:
- parameters = param_point
- keep = True
- if filter_conditions:
- keep = self.select_param_point(parameters, filter_conditions, default=True)
- if exclude_conditions:
- keep &= not (self.select_param_point(parameters, exclude_conditions, default=False))
- if keep:
- selected_param_points.append(param_point)
- return selected_param_points
+ def get_attrib(x):
+ if dict_keys is None:
+ return x
+ if isinstance(dict_keys, str):
+ return x[dict_keys]
+ if isinstance(dict_keys, list):
+ return dict(ChainMap(*[x[k] for k in dict_keys]))
+ raise ValueError('invalid value for "dict_keys"')
+ selected_points = param_points
+ if filter_expr is not None:
+ selected_points = [point for point in selected_points \
+ if self.is_point_selected(get_attrib(point), filter_conditions)]
+ if exclude_expr is not None:
+ selected_points = [point for point in selected_points \
+ if not self.is_point_selected(get_attrib(point), exclude_conditions)]
+ return selected_points
+
+ @staticmethod
+ def is_point_selected(param_point, conditions):
+ for condition in conditions:
+ selected = True
+ for param_name, match_functions in condition.items():
+ param_value = param_point[param_name]
+ selected &= any(match_func(param_value) for match_func in match_functions)
+ if selected:
+ return True
+ return False
def sanity_check(self, internal_param_points:Dict, external_param_points:Dict):
if not self.allow_none:
@@ -362,8 +370,8 @@ class ParamParser:
def get_param_points(self, dirname:str="",
filter_expr:Optional[str]=None,
exclude_expr:Optional[str]=None):
- external_param_points = self.get_external_param_points(dirname, filter_expr, exclude_expr)
- internal_param_points = self.get_internal_param_points(filter_expr, exclude_expr)
+ external_param_points = self.get_external_param_points(dirname)
+ internal_param_points = self.get_internal_param_points()
self.sanity_check(internal_param_points, external_param_points)
if len(internal_param_points) == 0:
internal_param_points = [{}]
@@ -379,4 +387,6 @@ class ParamParser:
param_point['external_parameters'] = {**ext_params}
param_point['internal_parameters'] = {**int_params}
param_points.append(param_point)
+ param_points = self.get_selected_points(param_points, filter_expr, exclude_expr,
+ dict_keys=['internal_parameters', 'external_parameters'])
return param_points
\ No newline at end of file
diff --git a/quickstats/plots/template.py b/quickstats/plots/template.py
index 38732aa08854151d1689870ad454fefe57d24ef4..fcd9ccef4fc373ebbc6558f4947fac8352f1604f 100644
--- a/quickstats/plots/template.py
+++ b/quickstats/plots/template.py
@@ -257,9 +257,10 @@ def ratio_frames(height_ratios:Tuple[int]=(3, 1), hspace:float=0.07,
if logy:
ax1.set_yscale('log')
-
- format_axis_ticks(ax1, x_axis=True, y_axis=True, x_axis_styles={"labelbottom":False},
- xtick_styles=styles['xtick'], ytick_styles=styles['ytick'], **styles['axis'])
+
+ ax1_styles = combine_dict(styles['axis'], {"x_axis_styles": {"labelbottom": False}})
+ format_axis_ticks(ax1, x_axis=True, y_axis=True, xtick_styles=styles['xtick'],
+ ytick_styles=styles['ytick'], **ax1_styles)
format_axis_ticks(ax2, x_axis=True, y_axis=True, xtick_styles=styles['xtick'],
ytick_styles=styles['ytick'], **styles['axis'])
@@ -701,7 +702,7 @@ def draw_analysis_label(axis, loc=(0.05, 0.95), fontsize:float=25, status:str='i
main_texts.extend(main_text.split("//"))
if colab is not None:
# add collaboration and status text
- colab_text = r"\bolditalic{" + colab + "} " + status_text
+ colab_text = r"\bolditalic{" + colab + "} " + status_text
main_texts.append(colab_text)
for text in main_texts:
_, _, ymin, _ = draw_text(axis, xmin, ymin, text,
diff --git a/quickstats/plots/upper_limit_2D_plot.py b/quickstats/plots/upper_limit_2D_plot.py
index 3fc25773609da381041f4b15b7a45ee387c38d3e..323486fca5225997a7f2aad9c362da3e3d36a177 100644
--- a/quickstats/plots/upper_limit_2D_plot.py
+++ b/quickstats/plots/upper_limit_2D_plot.py
@@ -8,7 +8,7 @@ import matplotlib.lines as lines
from quickstats.plots import AbstractPlot
from quickstats.utils.common_utils import combine_dict
-from quickstats.maths.interpolation import get_x_intersections
+from quickstats.maths.interpolation import get_intervals_between_curves
class UpperLimit2DPlot(AbstractPlot):
@@ -91,11 +91,14 @@ class UpperLimit2DPlot(AbstractPlot):
'constraint_options': {
'loc': (0.05, 0.25),
'dy': 0.06,
+ 'decimal_place': 2,
'use_signal_strength': True,
- 'expected_interval_fmt': r'Expected: {xlabel}$\in [{lower_interval:.2f}, {upper_interval:.2f}]$',
- 'observed_interval_fmt': r'Observed: {xlabel}$\in [{lower_interval:.2f}, {upper_interval:.2f}]$',
- 'expected_length_fmt': r'Allowed range: {length:.2f}',
- 'observed_length_fmt': r'Allowed range: {length:.2f}'
+ 'expected_interval_fmt': r'Expected: {xlabel}$\in [{lower_interval:.{decimal_place}f}'
+ r', {upper_interval:.{decimal_place}f}]$',
+ 'observed_interval_fmt': r'Observed: {xlabel}$\in [{lower_interval:.{decimal_place}f}'
+ r', {upper_interval:.{decimal_place}f}]$',
+ 'expected_length_fmt': r'Allowed range: {length:.{decimal_place}f}',
+ 'observed_length_fmt': r'Allowed range: {length:.{decimal_place}f}'
}
}
@@ -306,13 +309,15 @@ class UpperLimit2DPlot(AbstractPlot):
if (not use_signal_strength) and (scale_theory and (self.theory_func is not None)):
_, scale_y, _, _ = self.get_theory_prediction(x)
_, theo_y, _, _ = self.get_theory_prediction(theo_x)
- intersections = get_x_intersections(x, y * scale_y, theo_x, theo_y)
+ interval = get_intervals_between_curves(x, y * scale_y, theo_x, theo_y)
else:
- intersections = get_x_intersections(x, y, theo_x, np.ones(theo_x.shape))
- if len(intersections) != 2:
+ interval = get_intervals_between_curves(x, y, theo_x, np.ones(theo_x.shape))
+ if len(interval) == 0:
return None, None
- interval = intersections
- length = (intersections[1] - intersections[0])
+ if np.isfinite(interval).all():
+ length = (interval[1] - interval[0])
+ else:
+ length = None
return interval, length
def draw_constraint(self, ax, data:pd.DataFrame,
@@ -325,9 +330,10 @@ class UpperLimit2DPlot(AbstractPlot):
option:Optional[Dict]=None):
if option is None:
option = self.config['constraint_options']
+ decimal_place = option['decimal_place']
loc = option['loc']
dy = option['dy']
- txts = []
+ texts = []
x = data.index.values
for flag, column, key in [(draw_expected, '0', 'expected'),
(draw_observed, 'obs', 'observed')]:
@@ -336,15 +342,17 @@ class UpperLimit2DPlot(AbstractPlot):
y = data[column].values
interval, length = self.get_interval_and_length(x, y, scale_theory=scale_theory)
if draw_interval and (interval is not None):
- interval_txt = option[f'{key}_interval_fmt'].format(xlabel=xlabel,
+ interval_str = option[f'{key}_interval_fmt'].format(xlabel=xlabel,
lower_interval=interval[0],
- upper_interval=interval[1])
- txts.append(interval_txt)
+ upper_interval=interval[1],
+ decimal_place=decimal_place)
+ interval_str = interval_str.replace('-inf', r'N.A.').replace('inf', 'N.A.')
+ texts.append(interval_str)
if draw_length and (length is not None):
- length_txt = option[f'{key}_length_fmt'].format(length=length)
- txts.append(length_txt)
- for i, txt in enumerate(txts):
- ax.annotate(txt, (loc[0], loc[1] - i * dy), xycoords='axes fraction',
+ interval_str = option[f'{key}_length_fmt'].format(length=length, decimal_place=decimal_place)
+ texts.append(interval_str)
+ for i, text in enumerate(texts):
+ ax.annotate(text, (loc[0], loc[1] - i * dy), xycoords='axes fraction',
**self.styles['annotation'])
def _get_candidate_columns(self, data:pd.DataFrame):
diff --git a/quickstats/plots/variable_distribution_plot.py b/quickstats/plots/variable_distribution_plot.py
index c8d0d4944906b2292017a37f8ebe4480293addc2..826161d373f688492cdbf6eb79b166354273926c 100644
--- a/quickstats/plots/variable_distribution_plot.py
+++ b/quickstats/plots/variable_distribution_plot.py
@@ -62,7 +62,8 @@ class VariableDistributionPlot(AbstractPlot):
'error_label_format': r'{label} $\pm \sigma$',
'show_xerr': False,
'stacked_label': ':stacked_{index}:',
- 'box_legend_handle': False
+ 'box_legend_handle': False,
+ 'save_hist_data': False
}
def __init__(self, data_map:Union["pandas.DataFrame", Dict[str, "pandas.DataFrame"]],
@@ -121,6 +122,7 @@ class VariableDistributionPlot(AbstractPlot):
self.load_data(data_map)
self.plot_options = plot_options
self.label_map = label_map
+ self.reset_hist_data()
super().__init__(color_cycle=color_cycle,
styles=styles,
analysis_label_options=analysis_label_options,
@@ -287,7 +289,10 @@ class VariableDistributionPlot(AbstractPlot):
if ylim[1] < np.max(y):
ylim[1] = np.max(y)
ax.set_ylim(ylim)
-
+
+ if self.config['save_hist_data']:
+ self.hist_comparison_data.append(comparison_data)
+
return handle, error_handle
def deduce_bin_range(self, samples:List[str], column_name:str,
@@ -393,6 +398,10 @@ class VariableDistributionPlot(AbstractPlot):
for i, target in enumerate(plot_options):
self.update_legend_handles({target:handle[i]})
return bin_edges, hist_data
+
+ def reset_hist_data(self):
+ self.hist_data = {}
+ self.hist_comparison_data = []
def draw(self, column_name:str, weight_name:Optional[str]=None,
targets:Optional[List[str]]=None,
@@ -489,6 +498,7 @@ class VariableDistributionPlot(AbstractPlot):
bin_range = self.deduce_bin_range(relevant_samples, column_name, variable_scale=variable_scale)
self.stdout.info(f"Using deduced bin range ({bin_range[0]:.3f}, {bin_range[1]:.3f})")
+ self.reset_data()
binned_data = {}
target_bin_edges = {}
@@ -601,7 +611,10 @@ class VariableDistributionPlot(AbstractPlot):
self.decorate_comparison_axis(ax_ratio, **comparison_options)
ax.set(xlabel=None)
ax.tick_params(axis="x", labelbottom=False)
-
+
+ if save_hist_data:
+ self.hist_data = binned_data
+
if comparison_options is not None:
return ax, ax_ratio
diff --git a/quickstats/utils/common_utils.py b/quickstats/utils/common_utils.py
index d0d234aa1667b36c4cf7139f71c59576b4e7c290..2324c987a126083bcf6dbc3c5b35af14b55d8c32 100644
--- a/quickstats/utils/common_utils.py
+++ b/quickstats/utils/common_utils.py
@@ -12,7 +12,7 @@ import collections.abc
import numpy as np
-from .string_utils import split_str
+from .string_utils import split_str, parse_as_dict
def timely_info(green_text, normal_text, adjust=40):
print(datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S'), '\033[92m[INFO]\033[0m', '\033[92m{}\033[0m'.format(green_text).rjust(40, ' '), normal_text)
@@ -213,17 +213,44 @@ def get_class_that_defined_method(meth):
return cls
return getattr(meth, '__objclass__', None)
-def batch_makedirs(dirnames:Union[str, List[str], Dict[str, str]]):
+def batch_makedirs(dirnames:Union[str, List[str], Dict[str, str]]) -> None:
+ """
+ Create multiple directories in a batch.
+
+ This function accepts a single directory name as a string, a list of directory
+ names, or a dictionary where the values are directory names. It then creates
+ these directories if they do not already exist.
+
+ Parameters
+ ----------
+ dirnames : Union[str, List[str], Dict[str, str]]
+ The directory name(s) to be created.
+ - If it's a string, a single directory will be created.
+ - If it's a list, each element is treated as a directory name.
+ - If it's a dictionary, each value is treated as a directory name.
+
+ Raises
+ ------
+ ValueError
+ If `dirnames` is not a string, a list, or a dictionary.
+
+ Examples
+ --------
+ >>> batch_makedirs('new_dir')
+ >>> batch_makedirs(['new_dir1', 'new_dir2'])
+ >>> batch_makedirs({'dir1': 'new_dir1', 'dir2': 'new_dir2'})
+ """
if isinstance(dirnames, str):
- batch_makedirs([dirnames])
- elif isinstance(dirnames, list):
- for dirname in dirnames:
- if not os.path.exists(dirname):
- os.makedirs(dirname)
+ dirnames = [dirnames]
elif isinstance(dirnames, dict):
- for key in dirnames:
- if not os.path.exists(dirnames[key]):
- os.makedirs(dirnames[key])
+ dirnames = list(dirnames.values())
+
+ if isinstance(dirnames, list):
+ for dirname in dirnames:
+ abs_dirname = os.path.abspath(dirname)
+ os.makedirs(abs_dirname, exist_ok=True)
+ else:
+ raise ValueError('invalid format for "dirnames"')
def set_scripts_path(scripts_path, undo=False):
if (scripts_path in sys.path) and (undo==True):
@@ -386,4 +413,38 @@ def filter_dataframe_by_index_values(df, index_values:Union[Tuple[List], List],
index_levels = (index_levels,)
for values, level in zip(index_values, index_levels):
df = df.loc[df.index.get_level_values(level).isin(values)]
- return df
\ No newline at end of file
+ return df
+
+def parse_config_dict(expr:Optional[Union[str, Dict]]=None):
+ if expr is None:
+ return {}
+ if isinstance(expr, str):
+ return parse_as_dict(expr)
+ if isinstance(expr, dict):
+ return expr
+ raise ValueError(f'expr of type "{type(expr)}" is not convertible to a config dict')
+
+def update_config_dict(orig_dict, new_dict, allow_overlap_keys:bool=True):
+ orig_dict = parse_config_dict(orig_dict)
+ new_dict = parse_config_dict(new_dict)
+ if not allow_overlap_keys:
+ overlap_keys = list(set(orig_dict.keys()) & set(new_dict.keys()))
+ if overlap_keys:
+ raise RuntimeError(f'found overlap keys between two config dict: {", ".join(overlap_keys)}')
+ return combine_dict(orig_dict, new_dict)
+
+def list_of_dict_to_dict_of_list(source:List[Dict], use_first_keys:bool=True):
+ if use_first_keys:
+ return {k: [item[k] for item in source] for k in source[0]}
+ common_keys = set.intersection(*map(set, source))
+ return {k: [item[k] for item in source] for k in common_keys}
+
+def dict_of_list_to_list_of_dict(source:Dict[str, List]):
+ return [dict(zip(source, t)) for t in zip(*source.values())]
+
+def save_as_json(data:Dict, outname:str,
+ indent:int=2, truncate:bool=True):
+ with open(outname, "w") as file:
+ json.dump(data, file, indent=2)
+ if truncate:
+ file.truncate()
\ No newline at end of file
diff --git a/quickstats/utils/roofit_utils.py b/quickstats/utils/roofit_utils.py
index 6b1bacd0230edd3ea867fc1025cff25a4bde453f..497337c5294a7173f21733d6433eb850f6daf89b 100644
--- a/quickstats/utils/roofit_utils.py
+++ b/quickstats/utils/roofit_utils.py
@@ -42,7 +42,9 @@ def get_variable_attributes(variable:ROOT.RooAbsReal, asym_error:bool=False):
'value': variable.getVal()
}
else:
- raise ValueError("variable must be an instance of RooAbsReal")
+ attributes = {
+ 'name' : variable.GetName()
+ }
return attributes
def variable_collection_to_dataframe(var_collection:Union[list, ROOT.RooArgSet],
@@ -81,7 +83,7 @@ def construct_categorized_pdf_dataset(pdf:"ROOT.RooAbsPdf", dataset:"ROOT.RooDat
def factorize_pdf(observables:"ROOT.RooArgSet", pdf:"ROOT.RooAbsPdf", constraints:"ROOT.RooArgSet"):
pdf_class = pdf.ClassName()
- if pdf_class == "RooProdPdf":
+ if pdf_class.InheritsFrom("RooProdPdf"):
new_factors = ROOT.RooArgList()
new_owned = ROOT.RooArgSet()
pdf_list = pdf.pdfList()
@@ -108,7 +110,7 @@ def factorize_pdf(observables:"ROOT.RooArgSet", pdf:"ROOT.RooAbsPdf", constraint
factorized_pdf.addOwnedComponents(new_owned)
copy_attributes(pdf, factorized_pdf)
return factorized_pdf
- elif pdf_class == "RooSimultaneous":
+ elif pdf_class.InheritsFrom("RooSimultaneous"):
cat = pdf.indexCat().Clone()
n_bins = cat.numBins("")
factorized_pdfs = []
@@ -142,11 +144,10 @@ def factorize_pdf(observables:"ROOT.RooArgSet", pdf:"ROOT.RooAbsPdf", constraint
elif pdf.dependsOn(observables):
return pdf
else:
- if not constraints.contains(pdf):
+ if (not constraints.contains(pdf)) and (not pdf.getAttribute('ignoreConstraint')):
constraints.add(pdf)
return 0
-
def rebuild_simultaneous_pdf(observables:"ROOT.RooArgSet", sim_pdf:"ROOT.RooSimultaneous"):
assert sim_pdf.ClassName() == "RooSimultaneous"
constraints = ROOT.RooArgList()
diff --git a/quickstats/utils/roostats_utils.py b/quickstats/utils/roostats_utils.py
index 69091c72098bc6dc720d01d2fb3a9b66f3fadc95..43832d8dda3f0c219a374a2581d2d72f0164996e 100644
--- a/quickstats/utils/roostats_utils.py
+++ b/quickstats/utils/roostats_utils.py
@@ -8,6 +8,14 @@ from cppyy.gbl.std import vector
import ROOT
import ROOT.RooStats as RS
+from quickstats import DescriptiveEnum
+
+class PriorPdfType(DescriptiveEnum):
+ NONE = (0, "No prior")
+ FLAT = (1, "Uniform prior")
+ UNIFORM = (2, "Uniform prior")
+ INV_SQRT = (3, "Reciprocal of square root of POI")
+
def get_null_distribution(htr:ROOT.RooStats.HypoTestResult)->np.ndarray:
return np.array(htr.GetNullDistribution().GetSamplingDistribution().data())
@@ -213,4 +221,41 @@ def get_hypotest_data(htr):
cl_sb_err2 = pow(data['expected']['CLsplusbError'], 2)
data['expected']['CLsError'] = np.sqrt(cl_sb_err2 + cl_b_err2 * pow(data['expected']['CLs'], 2)) / data['expected']['CLb']
- return data
\ No newline at end of file
+ return data
+
+def set_prior_pdf(ws: "ROOT.RooWorkspace",
+ mc: "ROOT.RooStats.ModelConfig",
+ pois: "ROOT.RooArgSet",
+ prior_type:Optional[Union[PriorPdfType, str]]="flat",
+ prior_name:str="prior_pdf"):
+ if prior_type is None:
+ return
+ try:
+ prior_type = PriorPdfType.parse(prior_type)
+ except:
+ pass
+ if prior_type == PriorPdfType.NONE:
+ return
+ if prior_type in [PriorPdfType.UNIFORM, PriorPdfType.FLAT]:
+ prior_pdf = ROOT.RooUniform(prior_name, prior_name, pois)
+ ws.Import(prior_pdf)
+ mc.SetPriorPdf(prior_pdf)
+ return
+
+ def set_prior_from_expr(expr:str):
+ if len(pois) != 1:
+ raise RuntimeError('number of POIs must be 1 when using the inverse square root prior')
+ poi_name = pois.first().GetName()
+ prior_expr = f"EXPR::{prior_name}(\"{expr}\",{poi_name})"
+ ws.factory(prior_expr)
+ prior_pdf = ws.pdf(prior_name)
+ mc.SetPriorPdf(prior_pdf)
+
+ if prior_type == PriorPdfType.INV_SQRT:
+ set_prior_from_expr("1/sqrt(@0)")
+ elif isinstance(prior_type, str) and ("@0" in prior_type):
+ set_prior_from_expr(prior_type)
+ elif (prior_type and ws.pdf(prior_type)):
+ mc.SetPriorPdf(ws.pdf(prior_type))
+ else:
+ raise RuntimeError(f'unknown prior type: {prior_type}')
\ No newline at end of file
diff --git a/quickstats/utils/root_utils.py b/quickstats/utils/root_utils.py
index dc25cd395e5634bb7a6a8f4108dc976880ab9798..1da86d46f639677b557cb61a0f32275f5d85a379 100644
--- a/quickstats/utils/root_utils.py
+++ b/quickstats/utils/root_utils.py
@@ -377,4 +377,9 @@ def get_th1d_model(bins:Union[int, Sequence]=128,
xlow, xup = 0., 0.
else:
xlow, xup = bin_range
- return ROOT.RDF.TH1DModel(name, title, bins, xlow, xup)
\ No newline at end of file
+ return ROOT.RDF.TH1DModel(name, title, bins, xlow, xup)
+
+def close_read_cache(rfile:"ROOT.TFile"):
+ read_cache = rfile.GetCacheRead()
+ if read_cache:
+ read_cache.Close()
\ No newline at end of file
diff --git a/quickstats/utils/string_utils.py b/quickstats/utils/string_utils.py
index b160bbe3aa32e7b99186038d8ad435d9ebb7fe9c..ad4b535bd38875f9a4efc7f8c55bf282d1a8c953 100644
--- a/quickstats/utils/string_utils.py
+++ b/quickstats/utils/string_utils.py
@@ -1,5 +1,6 @@
-from typing import Optional, Callable
+from typing import Optional, Callable, List
import re
+import ast
def split_lines(s: str, comment_string: Optional[str] = "#", remove_blank: bool = True,
with_line_number: bool = False, keepends: bool = False):
@@ -37,7 +38,8 @@ def split_lines(s: str, comment_string: Optional[str] = "#", remove_blank: bool
return lines
-def split_str(s: str, sep: str = None, strip: bool = True, remove_empty: bool = False, cast: Optional[Callable] = None) -> list:
+def split_str(s: str, sep: str = None, strip: bool = True, remove_empty: bool = False, cast: Optional[Callable] = None,
+ use_paranthesis:bool = False) -> List:
"""
Splits a string and applies optional transformations.
@@ -54,18 +56,25 @@ def split_str(s: str, sep: str = None, strip: bool = True, remove_empty: bool =
the string is split at any whitespace. Defaults to None.
strip : bool, optional
Whether to trim leading and trailing whitespace from each substring. Defaults to True.
- remove_empty : bool, optional
+ remove_empty : bool, default = False
Whether to remove empty substrings from the list. Defaults to False.
cast : Callable, optional
An optional casting function to apply to each substring. It should be a function
that takes a single string argument and returns a value. Defaults to None.
+ use_paranthesis: bool, default = False
+ Whether to ignore separator within paranthesis.
Returns
-------
list
A list of substrings (or transformed substrings) obtained by splitting the input string.
"""
- items = s.split(sep)
+ if use_paranthesis:
+ if sep is None:
+ raise ValueError('separator can not be None when "use_paranthesis" option is set to True')
+ items = re.split(sep + r'\s*(?![^()]*\))', s)
+ else:
+ items = s.split(sep)
if strip:
items = [item.strip() for item in items]
if remove_empty:
@@ -74,6 +83,11 @@ def split_str(s: str, sep: str = None, strip: bool = True, remove_empty: bool =
items = [cast(item) for item in items]
return items
+def split_str_excl_paranthesis(s: str, sep: str = ",", strip: bool = True, remove_empty: bool = False) -> List:
+ regex = re.compile(sep + r'\s*(?![^()]*\))')
+
+
+
whitespace_trans = str.maketrans('', '', " \t\r\n\v")
newline_trans = str.maketrans('', '', "\r\n")
@@ -125,4 +139,41 @@ def remove_neg_zero(s:str):
print(remove_neg_zero(string))
# outputs: "The temperature is 0.000 degrees."
"""
- return neg_zero_regex.sub(r'\1', s)
\ No newline at end of file
+ return neg_zero_regex.sub(r'\1', s)
+
+
+def parse_as_dict(s:str, item_sep:str=',', key_value_sep:str='='):
+ """
+ Parse a string into a dictionary based on given item and key-value separators.
+
+ Parameters
+ ----------
+ s : str
+ The input string to be parsed into a dictionary.
+ item_sep : (optional) str, default = ','
+ The separator between items
+ key_value_sep : (optional) str, default = '='
+ The separator between keys and values
+
+ Returns
+ -------
+ dict
+ A dictionary containing the parsed key-value pairs.
+
+ Examples
+ --------
+ >>> parse_as_dict("name='John',age=25")
+ {'name': 'John', 'age': 25}
+ """
+ tokens = split_str(s, sep=item_sep, strip=True, remove_empty=True)
+ result = {}
+ for token in tokens:
+ subtokens = split_str(token, sep=key_value_sep, strip=True)
+ if len(subtokens) != 2:
+ raise ValueError(f'invalid key-value format: {token}')
+ key, value = subtokens
+ if key in result:
+ raise RuntimeError(f'multiple values specified for the key "{key}"')
+ result[key] = ast.literal_eval(value)
+ return result
+
\ No newline at end of file