V+jets NLO Madspin crash in gridpack

Asked by Andreas on 2018-02-12

Hi,

I am encountering a problem when trying to produce events from a gridpack that makes use of MadSpin.
I run 2.6.0 and the process is W+jets @ NLO with "decay V > q q" (also using bias weighting). Cards are at [1].

The initial run completes fine, but when I try to run from the gridpack, I get a crash (Logs at [2] and [3]).
Specifically, the crash happens on line 644 of madspin_interface.py, on this command:

    decay2['path'] = decay2['path'].replace(generate_all.path_me, self.options['ms_dir'])

It appears that this happens because one of the decay dicts in madspin.pkl has these contents:

    decay2 = {'matrix_element': None, 'finals': None, 'br': 1.1102230246251565e-16, 'base_order': None, 'path': None, 'decay_tag': None, 'decay_struct': None}

So basically 0 BR, but all other attributes set to None. If I add a simple check to avoid this crash, everything runs fine:

    if( not decay2['path'] ):
        continue
    else:
        decay2['path'] = decay2['path'].replace(generate_all.path_me, self.options['ms_dir'])

I also verified that the wanted decays of W->qq are present (printing path, branching and tag):

./madspingrid/full_me/SubProcesses/P2_uxcx_wmguxsx_wm_dux 0.3163456 ('wm_dux',)
./madspingrid/full_me/SubProcesses/P2_uxcx_wmguxsx_wm_dcx 0.01690461 ('wm_dcx',)
./madspingrid/full_me/SubProcesses/P2_uxcx_wmguxsx_wm_sux 0.01690412 ('wm_sux',)
./madspingrid/full_me/SubProcesses/P2_uxcx_wmguxsx_wm_scx 0.3163456 ('wm_scx',)
./madspingrid/full_me/SubProcesses/P2_uxcx_wmguxsx_wm_bux 3.802001e-06 ('wm_bux',)
./madspingrid/full_me/SubProcesses/P2_uxcx_wmguxsx_wm_bcx 0.0005602024 ('wm_bcx',)

Is it safe to just skip the pathologic decay or is this indicative of an underlying problem that needs to be understood?

Thanks

Andreas

[1] https://github.com/cmantill/genproductions/tree/mg26x/bin/MadGraph5_aMCatNLO/cards/production/13TeV/WJetsToQQ_5f_NLO_FXFX_bias
[2] Log of gridpack generation: http://cmantill.web.cern.ch/cmantill/gridpack_generationWJetsbias.log
[3] Log showing the error: http://cmantill.web.cern.ch/cmantill/cmsgrid_tag_1_debug.log

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Olivier Mattelaer
Solved:
2018-02-26
Last query:
2018-02-26
Last reply:
2018-02-21

Hi,

Yes this is perfectly save.

Note that such entry can actually be legitimate.
This correspond to the probability of loosing the events due to non uniform BR between different final states.
In your case you have W+ and W- decay and the BR difference is actually at the level of numerical uncertainty.
But if you put in your sample Z plus jets, then such contribution would be genuine.

But indeed, in that case they are not path/matrix element/... so your fix is fully save.

Thanks for the report,

Olivier

> On 12 Feb 2018, at 16:21, Andreas <email address hidden> wrote:
>
> New question #664319 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/664319
>
> Hi,
>
> I am encountering a problem when trying to produce events from a gridpack that makes use of MadSpin.
> I run 2.6.0 and the process is W+jets @ NLO with "decay V > q q" (also using bias weighting). Cards are at [1].
>
> The initial run completes fine, but when I try to run from the gridpack, I get a crash (Logs at [2] and [3]).
> Specifically, the crash happens on line 644 of madspin_interface.py, on this command:
>
> decay2['path'] = decay2['path'].replace(generate_all.path_me, self.options['ms_dir'])
>
> It appears that this happens because one of the decay dicts in madspin.pkl has these contents:
>
> decay2 = {'matrix_element': None, 'finals': None, 'br': 1.1102230246251565e-16, 'base_order': None, 'path': None, 'decay_tag': None, 'decay_struct': None}
>
> So basically 0 BR, but all other attributes set to None. If I add a simple check to avoid this crash, everything runs fine:
>
> if( not decay2['path'] ):
> continue
> else:
> decay2['path'] = decay2['path'].replace(generate_all.path_me, self.options['ms_dir'])
>
> I also verified that the wanted decays of W->qq are present (printing path, branching and tag):
>
> ./madspingrid/full_me/SubProcesses/P2_uxcx_wmguxsx_wm_dux 0.3163456 ('wm_dux',)
> ./madspingrid/full_me/SubProcesses/P2_uxcx_wmguxsx_wm_dcx 0.01690461 ('wm_dcx',)
> ./madspingrid/full_me/SubProcesses/P2_uxcx_wmguxsx_wm_sux 0.01690412 ('wm_sux',)
> ./madspingrid/full_me/SubProcesses/P2_uxcx_wmguxsx_wm_scx 0.3163456 ('wm_scx',)
> ./madspingrid/full_me/SubProcesses/P2_uxcx_wmguxsx_wm_bux 3.802001e-06 ('wm_bux',)
> ./madspingrid/full_me/SubProcesses/P2_uxcx_wmguxsx_wm_bcx 0.0005602024 ('wm_bcx',)
>
>
> Is it safe to just skip the pathologic decay or is this indicative of an underlying problem that needs to be understood?
>
> Thanks
>
> Andreas
>
> [1] https://github.com/cmantill/genproductions/tree/mg26x/bin/MadGraph5_aMCatNLO/cards/production/13TeV/WJetsToQQ_5f_NLO_FXFX_bias
> [2] Log of gridpack generation: http://cmantill.web.cern.ch/cmantill/gridpack_generationWJetsbias.log
> [3] Log showing the error: http://cmantill.web.cern.ch/cmantill/cmsgrid_tag_1_debug.log
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Andreas (balouderbaer) said : #2

Hi Olivier,

great, thanks for the feedback!

Andreas

Andreas (balouderbaer) said : #3

Thanks Olivier Mattelaer, that solved my question.