MadGraph5_aMC@NLO

Same events generated in successive runs

Asked by Matthew Low

Dear MadGraph team,

I have been generating events using cluster mode with SLURM. In multiple successive runs (i.e. run_01 and run_02 and run_03) the same events are being generated. In looking at the banners the rnd seed is changing, and the integrated weights are not completely identical:

< # Number of Events : 50000
< # Integrated weight (pb) : 0.46692139539
---
> # Number of Events : 0
> # Integrated weight (pb) : 0.466498528878

This may be coupled with another problem I've reported in the past about sometimes a run using SLURM failing at the very end of the run with the message:

^[[1;34mFail to read the number of unweighted events in the combine.log file^[[0m

The summary then reports zero events, but the unweighted_events.* files are populated. Can you suggested how to fix this?

Thanks,
- Matthew

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Olivier Mattelaer
Solved:
Last query:
Last reply:
Revision history for this message
Matthew Low (mattlow) said : 		#1

Just to add, this also happened when manually setting the rnd seed to different values.

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#2

Hi,

Did you have a central disk?
I guess that this related to a problem of finding the files from the node for that step.

Cheers,

Olivier
On Nov 8, 2014, at 1:07 AM, Matthew Low <email address hidden> wrote:

> Question #257011 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/257011
>
> Matthew Low gave more information on the question:
> Just to add, this also happened when manually setting the rnd seed to
> different values.
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Matthew Low (mattlow) said : 		#3

Hi,

Do you mean that perhaps the nodes are creating files that are not properly transferred over to the central disk so when the events are combined it is reading files from the previous run that were not overwritten by the current run?

- Matthew

Revision history for this message
Best Olivier Mattelaer (olivier-mattelaer) said : 		#4

Something like that yes.

Cheers,

Olivier
On Nov 24, 2014, at 6:06 AM, Matthew Low <email address hidden> wrote:

> Question #257011 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/257011
>
> Status: Answered => Open
>
> Matthew Low is still having a problem:
> Hi,
>
> Do you mean that perhaps the nodes are creating files that are not
> properly transferred over to the central disk so when the events are
> combined it is reading files from the previous run that were not
> overwritten by the current run?
>
> - Matthew
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Matthew Low (mattlow) said : 		#5

Thanks Olivier Mattelaer, that solved my question.