systList=list(),
)
if syst == "nominal":
categoryDict[catStr].sampleDict[sample].hist = obj
else:
if syst not in categoryDict[catStr].sampleDict[sample].systList:
categoryDict[catStr].sampleDict[sample].systList.append(
BaseObject(syst, hist=obj, vary=varyMap[vary]))
config = CardConfig(signal_name)
shapeStr = "shapes * * {fileName} $CHANNEL/$PROCESS $CHANNEL/$PROCESS_$SYSTEMATIC\n"
config.shapeStr = shapeStr.format(fileName=signal_name + "_shapes.root")
dataCard = DataCard(config)
cardDir = option.outputDir + "/" + dataCard.makeOutFileName("/", "")
reader = FileReader()
mkdir_p(option.outputDir)
writeObjDict = {}
binList = []
for catStr, cat in categoryDict.iteritems():
lnSystFile = lnSystElectronFile if 'isE' in catStr else lnSystMuonFile
shapeSystFile = shapeSystElectronFile if 'isE' in catStr else shapeSystMuonFile
bin = BaseObject(
catStr,
lnSystFile=lnSystFile,
shapeSystFile=shapeSystFile,
processList=list(),
signalNames=[
signal_name,
],
isSignal=lambda x: x == "sig",
systList=list(),
plot_ydim = 2
# ____________________________________________________________ ||
arr = np.load(input_csv_path)
arr_list, _ = preprocess_conditional_flow_data_cww(arr)
ndim = 3
ncond = 1
# ____________________________________________________________ ||
nf_model = ConditionalRealNVP(num_coupling_layers=10, ndim=ndim, ncond=ncond)
# ____________________________________________________________ ||
batch_trainer = MiniBatchTrainer()
nf_optimizer = tf.keras.optimizers.Adam()
mkdir_p(output_path)
for i_epoch in range(n_epoch):
idx_train = np.random.randint(0, len(arr_list), batch_size)
x_train = np.concatenate([arr_list[idx].x for idx in idx_train])
condition_train = np.concatenate(
[arr_list[idx].condition for idx in idx_train])
idx_batch = np.random.randint(0, x_train.shape[0], event_size)
x_train = x_train[idx_batch]
condition_train = condition_train[idx_batch]
nf_model.direction = -1
with tf.GradientTape() as tape:
nf_loss = nf_model.log_loss([x_train, condition_train])
grads = tape.gradient(nf_loss, nf_model.trainable_weights)
"etaL3",
"etaL4",
"phiL1",
"phiL2",
"phiL3",
"phiL4",
"idL1",
"idL2",
"idL3",
"idL4",
]
f = uproot.open(input_file_path)
t = f[input_tree_path]
model = tf.keras.models.load_model(model_path)
df = t.pandas.df(branches=branches)
x = np.reshape(df.massZ2.to_numpy() / scale, (1, df.shape[0]))
inputs = np.apply_along_axis(
lambda x: np.histogram(x, bins=bins, density=1.)[0], 1, x)
pred = model.predict(inputs)
ll = [
model.calculate_loss(pred, tf.constant([[b]], dtype=np.float32))[0]
for b in plot_bins
]
mkdir_p(plot_dir)
plt.plot(bins, ll)
plt.grid()
plt.savefig(os.path.join(plot_dir, plot_png))
inputBinName=binName,
signalNames=[
"Sig",
],
) for binName in binNames
if "-Norm" not in binName and "comb" not in binName
]
# ____________________________________________________________________________________________________________________________________________ ||
# syst
lnSystReader = LogNormalSystReader()
# ____________________________________________________________________________________________________________________________________________ ||
reader = FileReader()
mkdir_p(os.path.abspath(outputDir))
rateParamDict = {}
fakeRateDict = {}
paramList = []
sel_bin_list = []
for ibin, bin in enumerate(binList):
if option.verbose: print "-" * 20
if option.verbose: print bin.name
histName = bin.inputBinName
# bkg
for bkgName in bkg_names:
reader.openFile(inputDir, bkgName, TFileName)
print "Reading directory " + cardDir
inputFile = ROOT.TFile(
cardDir + "higgsCombineTest." + option.method + ".mH120.root", "READ")
tree = inputFile.Get("limit")
window_name = cardDir.split("/")[-2]
window_value = int(window_name.split("_")[1][1:])
if expOnly:
for i, entry in enumerate(tree):
outDict[quantiles[i]][window_value] = getattr(entry, varName)
else:
raise RuntimeError
# ________________________________________________________________ ||
# Draw limit with outDict
# ________________________________________________________________ ||
mkdir_p(os.path.dirname(option.outputPath))
nPoints = len(outDict["central"])
outGraphDict = {}
for plot in plots:
W = 800
H = 600
T = 0.08 * H
B = 0.12 * H
L = 0.12 * W
R = 0.04 * W
c = ROOT.TCanvas("c", "c", 100, 100, W, H)
if setLogY:
c.SetLogy()
c.SetFillColor(0)
c.SetBorderMode(0)
c.SetFrameFillStyle(0)
def init_job(self):
mkdir_p(self.output_path)