Is it possible to run proc_card in cluster mode?

Asked by Yannis Maznas on 2021-01-14

Dear MG5 developers,

I'm trying to create a new process directory using the following proc_card but the NLO part takes ages to finish even on multicore mode.
Is there a way to run this procedure in cluster mode?
I tried set run_mode = 1 but apparently it does not affect anything....

Best,
Yannis

=========================================================
set group_subprocesses Auto
set loop_optimized_output True
set low_mem_multicore_nlo_generation True
set loop_color_flows False
set max_npoint_for_channel 0
set ignore_six_quark_processes False
set gauge unitary
set complex_mass_scheme False
set nb_core 8
import model sm
define p = g u c d s u~ c~ d~ s~
define j = g u c d s u~ c~ d~ s~
define l+ = e+ mu+
define l- = e- mu-
define vl = ve vm vt
define vl~ = ve~ vm~ vt~
generate p p > j j mu+ mu- mu+ mu- QCD=0 QED=6 [QCD]
add process p p > j j e+ e- mu+ mu- QCD=0 QED=6 [QCD]
add process p p > j j e+ e- e+ e- QCD=0 QED=6 [QCD]
output VBS_nb_pp4ljj_SM_LT0_EWK_SM_NLO
=========================================================

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Olivier Mattelaer
Solved:
2021-01-14
Last query:
2021-01-14
Last reply:
2021-01-14

Hi,

First, do you need to have the muon/electron massive?
If not then you can certainly do
generate p p > j j l+ l- l+ l- QCD=0 QED=6 [QCD]
which should be close to three times faster.

Additionally, you can also use the mode
set low_mem_multicore_nlo_generation
which will activate the use of multicore for NLO generation (note thatt they are strong restriction on the version of python for that option)

Cheers,

Olivier

> On 14 Jan 2021, at 20:25, Yannis Maznas <email address hidden> wrote:
>
> New question #694995 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/694995
>
> Dear MG5 developers,
>
> I'm trying to create a new process directory using the following proc_card but the NLO part takes ages to finish even on multicore mode.
> Is there a way to run this procedure in cluster mode?
> I tried set run_mode = 1 but apparently it does not affect anything....
>
> Best,
> Yannis
>
> =========================================================
> set group_subprocesses Auto
> set loop_optimized_output True
> set low_mem_multicore_nlo_generation True
> set loop_color_flows False
> set max_npoint_for_channel 0
> set ignore_six_quark_processes False
> set gauge unitary
> set complex_mass_scheme False
> set nb_core 8
> import model sm
> define p = g u c d s u~ c~ d~ s~
> define j = g u c d s u~ c~ d~ s~
> define l+ = e+ mu+
> define l- = e- mu-
> define vl = ve vm vt
> define vl~ = ve~ vm~ vt~
> generate p p > j j mu+ mu- mu+ mu- QCD=0 QED=6 [QCD]
> add process p p > j j e+ e- mu+ mu- QCD=0 QED=6 [QCD]
> add process p p > j j e+ e- e+ e- QCD=0 QED=6 [QCD]
> output VBS_nb_pp4ljj_SM_LT0_EWK_SM_NLO
> =========================================================
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Yannis Maznas (imaznas) said : #2

Hi Olivier,

Thanks for your quick reply.
I think I do need them massive mostly for consistency purposes with older runs. (I'm not sure if massless e/μ will change anything crucial). I'll give i t a try nonetheless.

As already shown in the card posted, multicore mode was activated but with only 8 available cores the progress is slow.
Hence I was looking for a faster way (i.e cluster) though I suspect that's not an option at this step, right?

Cheers,
Yannis

Indeed no,

Now on cluster you easily have access to machine with more than 8 core available.
However the code will be limited in term of scalling.

Cheers,

Olivier

> On 14 Jan 2021, at 20:50, Yannis Maznas <email address hidden> wrote:
>
> Question #694995 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/694995
>
> Status: Answered => Open
>
> Yannis Maznas is still having a problem:
> Hi Olivier,
>
> Thanks for your quick reply.
> I think I do need them massive mostly for consistency purposes with older runs. (I'm not sure if massless e/μ will change anything crucial). I'll give i t a try nonetheless.
>
> As already shown in the card posted, multicore mode was activated but with only 8 available cores the progress is slow.
> Hence I was looking for a faster way (i.e cluster) though I suspect that's not an option at this step, right?
>
> Cheers,
> Yannis
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Yannis Maznas (imaznas) said : #4

Thanks Olivier Mattelaer, that solved my question.