composed_features['log_abs_l1_dz'] = lambda df: np.log10(np.abs(df.l1_dz))
composed_features['log_abs_l2_dxy'] = lambda df: np.log10(np.abs(df.l2_dxy))
composed_features['log_abs_l2_dz'] = lambda df: np.log10(np.abs(df.l2_dz))
composed_features['abs_q_01'] = lambda df: np.abs(df.hnl_q_01)
trainer = Trainer(
channel=ch,
base_dir=env['NTUPLE_DIR'],
#post_fix = 'HNLTreeProducer_%s/tree.root' %ch,
post_fix='HNLTreeProducer/tree.root',
features=[ #'l0_pt' ,
'l1_pt',
'l2_pt',
'hnl_dr_12',
'hnl_m_12',
'sv_prob',
'hnl_2d_disp',
],
composed_features=composed_features,
selection_data=selection,
selection_mc=selection + [cuts.selections['is_prompt_lepton']],
selection_tight=cuts.selections_pd['tight'],
lumi=59700.,
# epochs = 100,
# early_stopping = False,
)
if __name__ == '__main__':
trainer.train()
pass
selection = [
cuts.selections['pt_iso'],
cuts.selections['baseline'],
cuts.selections['vetoes_02_OS'],
cuts.selections['sideband'],
]
trainer = Trainer(
channel=ch + '_os',
base_dir=env['NTUPLE_DIR'],
#post_fix = 'HNLTreeProducer_%s/tree.root' %ch,
post_fix='HNLTreeProducer/tree.root',
features=[
'l0_pt',
'l1_pt',
'l2_pt',
'hnl_dr_12',
'hnl_m_12',
'sv_prob',
'hnl_2d_disp',
],
selection_data=selection,
selection_mc=selection + [cuts.selections['is_prompt_lepton']],
selection_tight=cuts.selections_pd['tight'],
lumi=59700.)
if __name__ == '__main__':
trainer.train()
pass