Manage memory during weighted events generating.

Asked by Anton Stepennov on 2018-11-04

Dear experts,

I'm trying to generate ~1000 weighted events with fixed order (NLO), with FO_ANALYSIS_FORMAT = LHE.
However, it takes too much memory (most in SubProcesses) which makes it impossible to do at cern lxplus, limited to 100GB,
is there any way (option), to handle these subprocesses alternately, so that there always was some memory available .

kind regards,
Anton

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Olivier Mattelaer
Solved:
2018-11-06
Last query:
2018-11-06
Last reply:
2018-11-04

Hi,

You are probably generating much more than 1k weighted events.
(Actually what you kind of choose is the number of weighted events PER CHANNEL of integration)

Note that you have one (hidden) parameter with which you can play in order to reduce the size of the computation:

fo_lhe_weight_ratio

This means that all weights below the maximum weight times that value will be (partly) unweighted to the weight max_weight*fo_lhe_weight_ratio
The default is 1e-3 but you can be more strict if needed.

Now I'm not sure that this will help you since this restriction happens quite late (at the time of the merging of all the weighted events generated). For the (I guess) complex process that you are looking at, it might be better to write your analysis in advance and not use that LHE format.

Cheers,

Olivier

On 4 Nov 2018, at 16:08, Anton Stepennov <<email address hidden><mailto:<email address hidden>>> wrote:

New question #675915 on MadGraph5_aMC@NLO:
https://answers.launchpad.net/mg5amcnlo/+question/675915

Dear experts,

I'm trying to generate ~1000 weighted events with fixed order (NLO), with FO_ANALYSIS_FORMAT = LHE.
However, it takes too much memory (most in SubProcesses) which makes it impossible to do at cern lxplus, limited to 100GB,
is there any way (option), to handle these subprocesses alternately, so that there always was some memory available .

kind regards,
Anton

--
You received this question notification because you are an answer
contact for MadGraph5_aMC@NLO.

Thanks Olivier Mattelaer, that solved my question.