def theta_limit_madminer(xsec=0.001,
lumi=1000000.0,
effect_phys=0.1,
effect_sys=0.1):
# Set up MadMiner file
miner = MadMiner()
miner.add_parameter(
lha_block="no one cares",
lha_id=12345,
parameter_name="theta",
morphing_max_power=1,
parameter_range=(-1.0, 1.0),
)
miner.add_benchmark({"theta": 0.0})
miner.add_benchmark({"theta": 1.0})
miner.set_morphing(include_existing_benchmarks=True, max_overall_power=1)
miner.save(".data.h5")
# Set up observations
proc = LHEProcessor(".data.h5")
proc.add_observable("x", "no one cares")
proc.reference_benchmark = "benchmark_0"
proc.nuisance_parameters = OrderedDict()
proc.nuisance_parameters["nu"] = ("benchmark_nuisance", None)
proc.observations = OrderedDict()
proc.observations["x"] = np.array([1.0])
proc.weights = OrderedDict()
proc.weights["benchmark_0"] = np.array([xsec])
proc.weights["benchmark_1"] = np.array([xsec * (1.0 + effect_phys)])
proc.weights["benchmark_nuisance"] = np.array([xsec * (1.0 + effect_sys)])
proc.save(".data2.h5")
# Calculate Fisher information
fisher = FisherInformation(".data2.h5")
info, cov = fisher.calculate_fisher_information_full_truth(theta=np.array(
[0.0]),
luminosity=lumi)
constraint = fisher.calculate_fisher_information_nuisance_constraints()
info = info + constraint
profiled = profile_information(info, [0])
# Uncertainty on theta
theta_limit = profiled[0, 0]**-0.5
# Remove file
os.remove(".data.h5")
os.remove(".data2.h5")
return theta_limit
if "madminer" not in key:
logging.getLogger(key).setLevel(logging.WARNING)
# Please enter here the path to your MG5 root directory. This notebook assumes that you installed Delphes and Pythia through MG5.
# In[3]:
mg_dir = '/home/software/MG5_aMC_v2_6_2/'
# ## 1. Generate events
# Let's load our setup:
# In[4]:
miner = MadMiner()
miner.load("data/setup.h5")
# In a next step, MadMiner starts MadGraph and Pythia to generate events and calculate the weights. You can use `run()` or `run_multiple()`; the latter allows to generate different runs with different run cards and optimizing the phase space for different benchmark points.
#
# In either case, you have to provide paths to the process card, run card, param card (the entries corresponding to the parameters of interest will be automatically adapted), and an empty reweight card. Log files in the `log_directory` folder collect the MadGraph output and are important for debugging.
#
# The `sample_benchmark` (or in the case of `run_all`, `sample_benchmarks`) option can be used to specify which benchmark should be used for sampling, i.e. for which benchmark point the phase space is optimized. If you just use one benchmark, reweighting to far-away points in parameter space can lead to large event weights and thus large statistical fluctuations. It is therefore often a good idea to combine at least a few different benchmarks for this option. Here we use the SM and the benchmark "w" that we defined during the setup step.
#
# One slight annoyance is that MadGraph only supports Python 2. The `run()` and `run_multiple()` commands have a keyword `initial_command` that let you load a virtual environment in which `python` maps to Python 2 (which is what we do below). Alternatively / additionally you can set `python2_override=True`, which calls `python2.7` instead of `python` to start MadGraph.
# In[5]:
miner.run(
sample_benchmark='sm',
mg_directory=mg_dir,
# In[2]:
# MadMiner output
logging.basicConfig(
format='%(asctime)-5.5s %(name)-20.20s %(levelname)-7.7s %(message)s',
datefmt='%H:%M',
level=logging.DEBUG)
# Output of all other modules (e.g. matplotlib)
for key in logging.Logger.manager.loggerDict:
if "madminer" not in key:
logging.getLogger(key).setLevel(logging.WARNING)
mg_dir = '/home/llr/cms/cortinovis/EFT2Obs/MG5_aMC_v2_6_7'
miner = MadMiner()
miner.load("/data_CMS/cms/cortinovis/ewdim6/data_ew_1M_az/setup.h5")
miner.run(
sample_benchmark='sm',
temp_directory='./',
mg_directory=mg_dir,
mg_process_directory=
'/data_CMS/cms/cortinovis/ewdim6/mg_processes_ew_1M_az/signal_pythia',
proc_card_file='cards/proc_card_signal.dat',
param_card_template_file='cards/param_card_template.dat',
pythia8_card_file='cards/pythia8_card.dat',
run_card_file='cards/run_card_signal_large.dat',
log_directory='/data_CMS/cms/cortinovis/ewdim6/logs_ew_1M_az/signal',
initial_command="PYTHONPATH=/usr/lib64/python",
python2_override=True,
#from matplotlib import pyplot as plt
#%matplotlib inline
import sys
import yaml
import inspect
from madminer.core import MadMiner
from madminer.plotting import plot_2d_morphing_basis
from madminer.sampling import combine_and_shuffle
from madminer.sampling import SampleAugmenter
from madminer.sampling import benchmark, benchmarks
from madminer.sampling import morphing_point, morphing_points, random_morphing_points
mg_dir = '/madminer/software/MG5_aMC_v2_6_2'
miner = MadMiner()#(debug=False)
input_file = str(sys.argv[1])
print('inputfile: ',input_file)
########### ADD parameters and benchmarks from input file
with open(input_file) as f:
# use safe_load instead load
dict_all = yaml.safe_load(f)
#get default values of miner.add_parameters()
default_arr = inspect.getargspec(miner.add_parameter)
default = dict(zip(reversed(default_arr.args), reversed(default_arr.defaults)))
from madminer.core import MadMiner
from madminer.plotting import plot_2d_morphing_basis
from madminer.sampling import combine_and_shuffle
from madminer.sampling import SampleAugmenter
input_file = str(sys.argv[1])
with open(input_file) as f:
dict_all = yaml.safe_load(f)
njobs = int(sys.argv[2])
h5_file = str(sys.argv[3])
mg_dir = '/home/software/MG5_aMC_v2_6_2'
miner = MadMiner() #(debug=False)
miner.load(h5_file)
##################################################################################
#signal
benchmarks = [str(i) for i in miner.benchmarks]
m = len(benchmarks)
print('list of benchmarks', benchmarks)
miner.run_multiple(
only_prepare_script=True,
#sample benchmarks from already stablished benchmarks in a democratic way
sample_benchmarks=benchmarks[0:njobs % m] + benchmarks * int(
# Please enter here the path to your MG5 root directory. This notebook assumes that you installed Delphes and Pythia through MG5.
# In[3]:
mg_dir = '/home/software/MG5_aMC_v2_6_2/'
# ## 1. Generate events
# Let's load our setup:
# In[4]:
miner = MadMiner()
miner.load("data/setup.h5")
delphes = DelphesReader('data/setup.h5')
# After creating the `DelphesReader` object, one can add a number of event samples (the output of running MadGraph and Pythia in step 1 above) with the `add_sample()` function.
#
# In addition, you have to provide the information which sample was generated from which benchmark with the `sampled_from_benchmark` keyword, and set `is_background=True` for all background samples.
# careful to keep same order of benchmarks as in the simulation file
benchmarks = ['sm', 'no-higgs','1.5pow4-higgs', '2pow4-higgs']
# Works up to 8 benchmarks
assert len(benchmarks) < 9
mg_dir = '/home/llr/cms/cortinovis/EFT2Obs/MG5_aMC_v2_6_7'
# Please enter here the path to your MG5 root directory. This notebook assumes that you installed Delphes and Pythia through MG5.
# In[248]:
# ## 1. Generate events
# Let's load our setup:
# In[250]:
miner = MadMiner()
miner.load("data_sme_hw_hbox/setup.h5")
# In a next step, MadMiner starts MadGraph and Pythia to generate events and calculate the weights. You can use `run()` or `run_multiple()`; the latter allows to generate different runs with different run cards and optimizing the phase space for different benchmark points.
#
# In either case, you have to provide paths to the process card, run card, param card (the entries corresponding to the parameters of interest will be automatically adapted), and an empty reweight card. Log files in the `log_directory` folder collect the MadGraph output and are important for debugging.
#
# The `sample_benchmark` (or in the case of `run_all`, `sample_benchmarks`) option can be used to specify which benchmark should be used for sampling, i.e. for which benchmark point the phase space is optimized. If you just use one benchmark, reweighting to far-away points in parameter space can lead to large event weights and thus large statistical fluctuations. It is therefore often a good idea to combine at least a few different benchmarks for this option. Here we use the SM and the benchmark "w" that we defined during the setup step.
#
# One slight annoyance is that MadGraph only supports Python 2. The `run()` and `run_multiple()` commands have a keyword `initial_command` that let you load a virtual environment in which `python` maps to Python 2 (which is what we do below). Alternatively / additionally you can set `python2_override=True`, which calls `python2.7` instead of `python` to start MadGraph.
# In[210]:
# Please enter here the path to your MG5 root directory. This notebook assumes that you installed Delphes and Pythia through MG5.
# In[321]:
mg_dir = '/home/llr/cms/cortinovis/EFT2Obs/MG5_aMC_v2_6_7'
# ## 1. Generate events
# Let's load our setup:
# In[323]:
miner = MadMiner()
miner.load("/home/llr/cms/cortinovis/miner/madminer/examples/tutorial_particle_physics/data_hel_w_b/setup.h5")
# In a next step, MadMiner starts MadGraph and Pythia to generate events and calculate the weights. You can use `run()` or `run_multiple()`; the latter allows to generate different runs with different run cards and optimizing the phase space for different benchmark points.
#
# In either case, you have to provide paths to the process card, run card, param card (the entries corresponding to the parameters of interest will be automatically adapted), and an empty reweight card. Log files in the `log_directory` folder collect the MadGraph output and are important for debugging.
#
# The `sample_benchmark` (or in the case of `run_all`, `sample_benchmarks`) option can be used to specify which benchmark should be used for sampling, i.e. for which benchmark point the phase space is optimized. If you just use one benchmark, reweighting to far-away points in parameter space can lead to large event weights and thus large statistical fluctuations. It is therefore often a good idea to combine at least a few different benchmarks for this option. Here we use the SM and the benchmark "w" that we defined during the setup step.
#
# One slight annoyance is that MadGraph only supports Python 2. The `run()` and `run_multiple()` commands have a keyword `initial_command` that let you load a virtual environment in which `python` maps to Python 2 (which is what we do below). Alternatively / additionally you can set `python2_override=True`, which calls `python2.7` instead of `python` to start MadGraph.
# In[344]:
additional_benchmarks = ['w', 'neg_w', 'ww', 'neg_ww']