MadDM

parallel running maddm

Asked by Qibin Liu

Hi experts,
    I hope to know how to run the maddm scan in parallel. I set mg5_configuration with the "nb_core" setting but it seems the maddm still run as one thread.
    I have tried a naive parallel method like submit parallel jobs but finally many jobs failed with errors. After investigation, I noticed when running the maddm and loading a model, it will automatically call the function "update_model_with_EFT" and then add a merged model like "xxx_REAL" in the madgraph install folder. This operation is called with "--recreate" argument thus it removes and recreates the common files and this prevent parallel running of the maddm. (e.g. two independent process will delete/write the same files).
    So I was wondering what is the recommended way to run the maddm in a parallel environment, like when I do a scan in a multi-dimension parameter space of a model to test the relic density and so on.

    The maddm version I use is v3.1.0
Thanks!

Question information

Language:
English Edit question
Status:
Answered
For:
MadDM Edit question
Assignee:
No assignee Edit question
Last query:
Last reply:
Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#1

Hi,

In that case, I would start by creating one directory.
Then do a cp -rf
the amount of time that you want to parrallize it
then use the internal scan method of the param_card to scan within each of them.

Cheers,

Olivier

> On 27 Mar 2021, at 15:20, Qibin Liu <email address hidden> wrote:
>
> New question #696275 on MadDM:
> https://answers.launchpad.net/maddm/+question/696275
>
> Hi experts,
> I hope to know how to run the maddm scan in parallel. I set mg5_configuration with the "nb_core" setting but it seems the maddm still run as one thread.
> I have tried a naive parallel method like submit parallel jobs but finally many jobs failed with errors. After investigation, I noticed when running the maddm and loading a model, it will automatically call the function "update_model_with_EFT" and then add a merged model like "xxx_REAL" in the madgraph install folder. This operation is called with "--recreate" argument thus it removes and recreates the common files and this prevent parallel running of the maddm. (e.g. two independent process will delete/write the same files).
> So I was wondering what is the recommended way to run the maddm in a parallel environment, like when I do a scan in a multi-dimension parameter space of a model to test the relic density and so on.
>
> The maddm version I use is v3.1.0
> Thanks!
>
> --
> You received this question notification because your team maddm is an
> answer contact for MadDM.

Can you help with this problem?

Provide an answer of your own, or ask Qibin Liu for more information if necessary.

To post a message you must log in.