"INFO: fail to reach target" in MG5_v2.9.9

Asked by Saumyen Kundu

Hi Olivier,
I am trying to generate events for the process 'Z+jets'or 'W+jets' in the multicore mode in MG5_v2.9.9 through multi_run with 60k events in each run. But it's failing to generate 60k events in most of the cases. The generated events though are not the same in each run varying from 1k to 50k, sometimes 60k also. The script I'm using for process generation is in below:
----------------
generate p p > z, z > vl vl~ @1
add process p p > z j, z > vl vl~ @2
add process p p > z j j, z > vl vl~ @3
add process p p > z j j j, z > vl vl~ @4
add process p p > z j j j j, z > vl vl~ @5
output
----------------
I found some posts on these. Like, in
https://answers.launchpad.net/mg5amcnlo/+question/691494

you have suggested setting 'hard_survey'>0 and in
https://answers.launchpad.net/mg5amcnlo/+question/700705

you have suggested setting 'sde_strategy'>0. But I noticed that "set sde_strategy 2" doesn't work have to be set manually inside the run_card. Should I use "set hard_survey 2" and "set job_strategy 1" to reduce the CPU time?

Thanks a lot for any help with this.

Regards,
Saumyen

Question information

Language:
English Edit question
Status:
Expired
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Last query:
Last reply:
Revision history for this message
Saumyen Kundu (saumyen.k) said :
#1

Sorry, "set SDE_strategy 2" worked. But on giving this it says "INFO: modify parameter SDE_strategy of the run_card.dat to None "

Regards,

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#3

Here one option to simplify the computation is likely to not decay the Z and let such decay be handle by the parton-shower.
Or if spin-correlation are critical for you (I doubt with decay to neutrino) you can use MadSpin for such decay (which is likely more efficient)

Cheers,

Olivier

Revision history for this message
Saumyen Kundu (saumyen.k) said :
#4

Thanks a lot, Olivier.

>Here one option to simplify the computation is likely to not decay the Z and let such decay be handled by the parton-shower.

I don't know how to do that. Is it that I leave it as "generate p p > z j j" and do the rest as it is. And at the time of parton shower, it will automatically be decayed invisibly through QED shower? Or something else I have to do?

Also for the process
------
define ww = w- w+; define bb = b b~; define ll = l- l+; define vll = vl vl~
generate p p > ww, ww > ll vll @1
add process p p > ww j, ww > ll vll @2
add process p p > ww j j, ww > ll vll @3
add process p p > ww j j j, ww > ll vll @4
add process p p > ww j j j j, ww > ll vll @5
output wjj
-----
It is taking a huge time (even more than 24 hours) for compilation particularly the process with 4 quarks.
----
INFO: Compiling for process 6/19.
INFO: P5_qq_wpqqqq_wp_lvl
----
Is there a way to reduce the CPU time? I am running via for-loop., although 'multi_run' is also working. And I using
----
set SDE_strategy 2
set hard_survey 1
set job_strategy 1
----

Regards,
Saumyen

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#5

> I don't know how to do that. Is it that I leave it as "generate p p > z j j" and do the rest as it is. And at the time of parton shower, it will automatically be decayed invisibly through QED shower? Or something else I have to do?

You will need to specify to the parton-shower that you want the invisible decay (check the manual of the parton-shower that you use)
For the rest yes, you can do it like this
generate p p > z @1
add process p p > z j @2
add process p p > z j j @3
add process p p > z j j j @4
add process p p > z j j j j @5

> Is there a way to reduce the CPU time?

Yes they are ways to fasten the compilation time, but this will remove optimization of the code and therefore the next step will be slower...
You can play (i.e. turn off) the optimization from this paper: 2102.00773 [hep-ph]

Cheers,

Olivier

Revision history for this message
Saumyen Kundu (saumyen.k) said :
#6

Thanks a lot, Olivier.
So, I tried the process (upto 3 jets) in two ways, decaying the Z-boson in MG5 itself as well as through Pythia8. For decaying through PY8, I added the following lines in the PY8 card, (hope this is the correct way)
----
23:onMode=off
23:onIfAny=12 14 16
----
But I see that the matched xsections and vetoed events are different for the 2 cases: 144.5pb (1418 events) while decaying in MG5 and 4.619e+04pb (3479 events) while decaying in PY8. Are these values expected? I was expecting they should be similar given the fact that the hadron level information in the HepMC files should be the same.

Regards,
Saumyen

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#7

Hi,

If I do:

generate p p > z, z > vl vl~ @1
output
launch
shower=PY8

I get:
     Cross-section : 8412 +- 12.06 pb
     Nb of events : 10000

If I do:
generate p p > z, z > vl vl~ @1
add process p p > z j, z > vl vl~ @2
output
launch
shower=PY8

I get:
  === Results Summary for run: run_01 tag: tag_1 ===

     Cross-section : 1.009e+04 +- 17.46 pb
     Nb of events : 10000
     Pythia8 merged cross-sections are:
      > Merging scale = 45 : 8425.9 +/- 9.2 [pb]
      > Merging scale = 67.5 : 8437 +/- 9.2 [pb]
      > Merging scale = 90 : 8445 +/- 9.2 [pb]

So I do not reproduce your numbers for the first one (and have number certainly closer to your decay within Pythia8 --even if I still do have a factor of two).

Revision history for this message
Launchpad Janitor (janitor) said :
#8

This question was expired because it remained in the 'Open' state without activity for the last 15 days.