Fail to combine runs/events

Asked by Eirini Kasimi on 2019-11-24

Hi,

I'm trying to run 100k events of the following process with the Eboli model:

generate p p > e- e+ mu- vm~ j j QCD<=2 T0==0 T1==0

It runs for two days and after the combing events step and the final cross section, it says that fail to combine events and the run ends with 21k events. The LHE file shows that has 100k events but the ntuple says 21k. When I try to run 10k events, it runs without problems. Do you have any idea about this?

Thanks in advance,
Eirini Kasimi

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Olivier Mattelaer
Solved:
2019-11-24
Last query:
2019-11-24
Last reply:
2019-11-24

What do you mean by the ntuple?
I do not see any output close to that.
If you use an external code to create such tuple, you should contact the authors of such tools.

Cheers,

Olivier

> On 24 Nov 2019, at 16:32, Eirini Kasimi <email address hidden> wrote:
>
> New question #686190 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/686190
>
> Hi,
>
> I'm trying to run 100k events of the following process with the Eboli model:
>
> generate p p > e- e+ mu- vm~ j j QCD<=2 T0==0 T1==0
>
> It runs for two days and after the combing events step and the final cross section, it says that fail to combine events and the run ends with 21k events. The LHE file shows that has 100k events but the ntuple says 21k. When I try to run 10k events, it runs without problems. Do you have any idea about this?
>
> Thanks in advance,
> Eirini Kasimi
>
>
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Eirini Kasimi (eirinikasimi) said : #2

 When the madgraph ends and writes the infos of the run, the cross section and the number of events are displayed on the screen. The number of events is not 100k events that i asked but 21k and above the results it said that fail to combine the events but this is not a problem when I run 10k events. So, madgraph is not run properly.

Thanks,
Eirini

Ok,

Then in that case, it seems that your efficiency for your process/model is too low to allow us to reach 100k events. In principle you can increase the number of allowed iteration to reach the accuracy but the simplest method is likely to just run
./bin/madevent multi_run 10
and ask that each of those 10 runs have nevents=10k

At the end the 10 lhe files will be merged into a single one.

Cheers,

Olivier

> On 24 Nov 2019, at 16:52, Eirini Kasimi <email address hidden> wrote:
>
> Question #686190 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/686190
>
> Eirini Kasimi posted a new comment:
> When the madgraph ends and writes the infos of the run, the cross
> section and the number of events are displayed on the screen. The number
> of events is not 100k events that i asked but 21k and above the results
> it said that fail to combine the events but this is not a problem when I
> run 10k events. So, madgraph is not run properly.
>
> Thanks,
> Eirini
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Eirini Kasimi (eirinikasimi) said : #4

Thanks Olivier Mattelaer, that solved my question.

Eirini Kasimi (eirinikasimi) said : #5

Hi Olivier,

I would like to ask you something more. You said that ", it seems that your efficiency for your process/model is too low to allow us to reach 100k events" but when i try to generate the following process with the eboli model, it can produce 100k events:

generate p p > e- e+ mu- vm~ j j T1=1 T1=1

So, in that case why don't we have the same problem?

Hi,

Well those are not the same matrix-element.

Such dimension 8 (or 6?) operator typically breaks unitarity very quickly.
and even if you keep your coupling small enough to avoid that issue, such operator higly enhanced the large energy contribution of the process.
So even if they are very similar in syntax to your other process they are quite different for the phase-space integration and therefore this is not surprising that you do not have the same issue for the two processes.

Additionally, since you have more Feynman Diagram, you technically have a completely different strategy of integration and even if such diagram are technically negligeable, the associated channel can still help to improve the convergence of the other channel of integration since the presence of interference makes them to have a non trivial interplay in term of global efficiency.

Cheers,

Olivier

> On 25 Nov 2019, at 10:02, Eirini Kasimi <email address hidden> wrote:
>
> Question #686190 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/686190
>
> Eirini Kasimi posted a new comment:
> Hi Olivier,
>
> I would like to ask you something more. You said that ", it seems that
> your efficiency for your process/model is too low to allow us to reach
> 100k events" but when i try to generate the following process with the
> eboli model, it can produce 100k events:
>
> generate p p > e- e+ mu- vm~ j j T1=1 T1=1
>
> So, in that case why don't we have the same problem?
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.