Gridpack NaN Cross Section

Hello
I had a more indepth look at the gridpack... only one process had problems. Its the process P4_qq_llqqqq/G579. I copied the content of the directory here:

/afs/cern.ch/work/s/santocch/public/DYJets_HT-extralow_mll/P4_qq_llqqqq_G579

Any help/advice you can give me is really appreciate

Best

Attilio

Hi,

Indeed I do not have any lxplus account, if you can put that information in a public space, this would be useful.

Cheers,

Olivier

> On 21 Jun 2017, at 23:14, attilio santocchia <email address hidden> wrote:
>
> Question #644315 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/644315
>
> attilio santocchia gave more information on the question:
> Hello
> I had a more indepth look at the gridpack... only one process had problems. Its the process P4_qq_llqqqq/G579. I copied the content of the directory here:
>
> /afs/cern.ch/work/s/santocch/public/DYJets_HT-
> extralow_mll/P4_qq_llqqqq_G579
>
> Any help/advice you can give me is really appreciate
>
> Best
>
> Attilio
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

No problem. You can find both the gridpack and the specific folder where the NaN appears here:

https://www.dropbox.com/sh/hx05ta179ld5dl3/AABhoeR_Q1kUCU3Z2Eb6cjzNa?dl=0

Thanks for helping

Attilio

Hi,

Your gridpack is not complete (all .f files are missing), can you create one with those files preserved?
This is impossible for me to debug without the possibility to recompile.

Cheers,

Olivier

> On 22 Jun 2017, at 10:47, attilio santocchia <email address hidden> wrote:
>
> Question #644315 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/644315
>
> attilio santocchia posted a new comment:
> No problem. You can find both the gridpack and the specific folder where
> the NaN appears here:
>
> https://www.dropbox.com/sh/hx05ta179ld5dl3/AABhoeR_Q1kUCU3Z2Eb6cjzNa?dl=0
>
> Thanks for helping
>
> Attilio
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi
In CMS we have a script that prepare the gridpack and delete all the unneeded files... I guess I can edit that script to keep all the files... but it will take a while. I'll be back as soon as I have the full gridpack available (I guess few days).

Thanks

Attilio

Hi,
I hope that my "naive" approach is correct... reproducing the problematic gridpack with all the files is a very long job so I thought that the files you need (the .f files) are the seme regardless if I produce Drell-Yan events with 1 jets or 4 jets (the former is not giving any problem and it's quite quick and the latter is very long and it gives the famous NaN xSection). So that is what I did... and you can find in the same dropbox link the pilotrun_gridpack.tar.gz file where in principle you have all the *.f files you need...

To summarize: the new file is the complete gridpack (no files deleted) from the same run_card.dat and a reduced proc_card.dat.
The original proc_card.dat was:
generate p p > l+ l- / h @0
add process p p > l+ l- j / h @1
add process p p > l+ l- j j / h @2
add process p p > l+ l- j j j / h @3
add process p p > l+ l- j j j j / h @4

The proc_card.dat I used has just:
generate p p > l+ l- / h @0
add process p p > l+ l- j / h @1

Am I too naive or is this what you need for trying to reproduce the problem?

Cheers

Attilio

Unfortunately this is not enough.

On the other hand, you can do the following:
define q = p / g
> generate q q > l+ l- q q q q / h @4

This should allow the generation to be already much faster.

You can also trick the code to generate only the G579
For this you have to edit
1) madgraph/madevent/gen_ximprove.py

You have to remove all the following lines:
          # Launch gensym
            p = misc.Popen(['./gensym'], stdout=subprocess.PIPE,
                                 stderr=subprocess.STDOUT, cwd=Pdir)
            #sym_input = "%(points)d %(iterations)d %(accuracy)f \n" % self.opts
            (stdout, _) = p.communicate('')

            if os.path.exists(pjoin(self.me_dir,'error')):
                files.mv(pjoin(self.me_dir,'error'), pjoin(Pdir,'ajob.no_ps.log'))
                P_zero_result.append(subdir)
                continue

            jobs = stdout.split()
And replace those by
     jobs = ['579']

2) SubProcesses/P4_qq_llqqqq/symfact.dat
And remove ALL lines but the ones with 579 (Note that you can not have empty line on that file and therefore you can not have any return symbol on the first line)
If that file does not exists, then you can copy the file symfact_orig.dat to symfact.dat and then doing it.

I never tried the above for gridpack but it should work as well.
Like this you will create a gridpack for only the channel with NAN that you report

Cheers,

Olivier

> On 22 Jun 2017, at 15:04, attilio santocchia <email address hidden> wrote:
>
> Question #644315 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/644315
>
> attilio santocchia posted a new comment:
> Hi,
> I hope that my "naive" approach is correct... reproducing the problematic gridpack with all the files is a very long job so I thought that the files you need (the .f files) are the seme regardless if I produce Drell-Yan events with 1 jets or 4 jets (the former is not giving any problem and it's quite quick and the latter is very long and it gives the famous NaN xSection). So that is what I did... and you can find in the same dropbox link the pilotrun_gridpack.tar.gz file where in principle you have all the *.f files you need...
>
> To summarize: the new file is the complete gridpack (no files deleted) from the same run_card.dat and a reduced proc_card.dat.
> The original proc_card.dat was:
> generate p p > l+ l- / h @0
> add process p p > l+ l- j / h @1
> add process p p > l+ l- j j / h @2
> add process p p > l+ l- j j j / h @3
> add process p p > l+ l- j j j j / h @4
>
> The proc_card.dat I used has just:
> generate p p > l+ l- / h @0
> add process p p > l+ l- j / h @1
>
> Am I too naive or is this what you need for trying to reproduce the
> problem?
>
> Cheers
>
> Attilio
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hello,
I have tricked the code to generate only the G579 following your instructions and this time the process didn't end with NaN but the cross-section was calculated correctly. I copied the file pilotrun_gridpack.tar.gz on the same dropbox folder (I renamed the previous file to old_pilotrun_gridpack.tar.gz). Just to double-check I'm just running again the original gridpack generation... to see if the NaN problem can be reproduced...

Cheers

Attilio

Hi
I run again the full generation and I believe I was able to keep all files... the xsection is again NaN and the problematic process is the same. The file is again in https://www.dropbox.com/sh/hx05ta179ld5dl3/AABhoeR_Q1kUCU3Z2Eb6cjzNa?dl=0

Cheers

Attilio

Hi Oliver,
I added the line to the run_card.dat and it seems that the problem is solved.
Thanks

Attilio

Thanks Olivier Mattelaer, that solved my question.