MadGraph5_aMC@NLO

how to change random seed between multi_run

Asked by jmli

Dear authors,

I want to run MG5 with multi_run on two computers.My question is how can I get these two multi_run with different results?
I have tried to change the iseed in run_card.dat, but MG5 change it to 0 after it begins to multi_run.

What's more, where can I find the actual seed which MG5 used?

Thanks in advance.
Jinmian

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Johan Alwall
Solved:
Last query:
Last reply:
Revision history for this message
Johan Alwall (johan-alwall) said : 		#1

Hello Jinmian,

MadGraph uses the seed in the run_card to get the starting seed for the multi_run.

In previous versions, if you gave a run_card seed in a multirun, all runs would be identical, since they used the same seed (as given in the run_card).

In recent versions, we have fixed this by using the seed in the run_card to set the initial seed, and then update the seed between runs.

The seed is updated in steps of 9 for regular runs (3 each for the survey and the two refines), so a safe way to use multiple computers for multi_runs is either to use seed 1 for the first computer, 2 for the second, etc., up to a maximum of 9. If you need more computers, please use higher seeds (at least 10 + Nrun*9 higher than the lower seeds).

You can make absolutely sure that no seed is used for multiple runs by grep:ing for iseed in the resulting banners of all your runs, and make sure they are all different.

All the best,
Johan

Revision history for this message
jmli (phyljm) said : 		#2

Hi Johan,
Thanks for your reply.
But I still have the problem:

I am using MG5_v1_5_3. In my process directory, I change the iseed in run_card.dat to 20. Then I run ./bin/madevent multi_run 20.
After this, I find the iseed in run_card.dat changes to 0 automatically. And In my run_01_0_tag_1_banner.txt, the iseed is also shows as 0.

Best,
Jinmian

Revision history for this message
Best Johan Alwall (johan-alwall) said : 		#3

Hello Jinmian,

In v. 1.5.3, the seed in the banner is not correctly updated when Pythia is run (it does show correctly in the unweighted_events.lhe file header). This has been fixed in v. 1.5.4.

So your runs should be ok, just that the seed doesn't show in the banner. Note that the seed in the run_card *is supposed to* be zero after the start of the run, as I explained above.

All the best,
Johan

Revision history for this message
jmli (phyljm) said : 		#4

Thanks Johan Alwall, that solved my question.