Showering does not work when running with nb_core=1

Bug #1716425 reported by Daniel Egana-Ugrinovic on 2017-09-11
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
MadGraph5_aMC@NLO
Undecided
Unassigned

Bug Description

Hi,

When showering with Pythia8 with just one core, showering fails. The error one gets is

Pythia8 cross-section could not be retreived.
Try turning parallelization off by setting the option nb_core to 1.

The given message doesn't make sense, since I am actually running on a single core. Strangely, the problem only arises when working with only one core, with generate_events --nb_core=1. Using generate_events --nb_core=2 fixes the issue (unfortunately, when running on clusters, system administrators usually ask you to avoid multi-threaded code, so in those cases using nb_core=2 would not help). The problem may be that the file unweighted_events.lhe is not being compressed into a gzip file unweighted_events.lhe.gz. Since run_shower.sh seems to need a gzipped lhe file, showering fails.

Further information:
Options are set in mg5 with
set run_mode 2
set nb_core 1
save options

Events are being run with madevent:
generate_events runname --nb_core=1
shower=PYTHIA8

Technical details:
Running madgraph 2.6.0
Pythia8
Python 2.7.13
gcc (GCC) 6.2.0

I attach the MG5 output file

Thanks!

Daniel

description: updated
description: updated
Download full text (4.5 KiB)

Thanks Daniel, I will take a look probably tommorow or on Wednesday.

Cheers,

Olivier

> On 11 Sep 2017, at 18:56, Daniel Egana-Ugrinovic <email address hidden> wrote:
>
> Public bug reported:
>
> Hi,
>
> When showering with Pythia8 with just one core, showering fails. The
> error one gets is
>
> Pythia8 cross-section could not be retreived.
> Try turning parallelization off by setting the option nb_core to 1.
>
> The given message doesn't make sense, since I am actually running on a
> single core. Strangely, the problem only arises when working with only
> one core, with generate_events --nb_core=1. Using generate_events
> --nb_core=2 fixes the issue (unfortunately, when running on clusters,
> system administrators usually ask you to avoid multi-threaded code, so
> in those cases using nb_core=2 would not help). The problem may be that
> the file unweighted_events.lhe is not being compressed into a gzip file
> unweighted_events.lhe.gz. Since run_shower.sh seems to need a gzipped
> lhe file, showering fails.
>
> Further information:
> Options are set in mg5 with
> set run_mode 2
> set nb_core 1
> save options
>
> Events are being run with madevent:
> generate_events runname --nb_core=1
> shower=PYTHIA8
>
> Technical details:
> Running madgraph 2.6.0
> Pythia8
> Python 2.7.13
> gcc (GCC) 6.2.0
>
> I attach the MG5 output file
>
> Thanks!
>
> Daniel
>
> ** Affects: mg5amcnlo
> Importance: Undecided
> Status: New
>
> ** Attachment added: "MG5outputrun_yL_0.5_DM_2.0_MFc_60.0_Mphi_110.0_run0"
> https://bugs.launchpad.net/bugs/1716425/+attachment/4948030/+files/MG5outputrun_yL_0.5_DM_2.0_MFc_60.0_Mphi_110.0_run0
>
> ** Description changed:
>
> Hi,
>
> When showering with Pythia8 with just one core, showering fails. The
> error one gets is
>
> Pythia8 cross-section could not be retreived.
> Try turning parallelization off by setting the option nb_core to 1.
>
> - The given message doesn't make sense, since I am actually running on
> - single core mode. Strangely, the problem only arises when working with
> - only one core, with generate_events --nb_core=1. Using generate_events
> + The given message doesn't make sense, since I am actually running on a
> + single core. Strangely, the problem only arises when working with only
> + one core, with generate_events --nb_core=1. Using generate_events
> --nb_core=2 fixes the issue (unfortunately, when running on clusters,
> system administrators usually ask you to avoid multi-threaded code, so
> in those cases using nb_core=2 would not help).
>
> Further information:
> Options are set in mg5 with
> set run_mode 2
> set nb_core 1
> save options
>
> Events are being run with madevent:
> generate_events runname --nb_core=1
> - shower=PYTHIA8
> + shower=PYTHIA8
>
> Technical details:
> Running madgraph 2.6.0
> Pythia8
> Python 2.7.13
> gcc (GCC) 6.2.0
>
> I attach the MG5 output file
>
> Thanks!
>
> Daniel
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/1716425
>
> Title:
> Showering does not work when running with nb_core=1
>
> Status in MadGraph5_aMC@NLO:
>...

Read more...

Download full text (3.3 KiB)

Hi Daniel,

Ok looks like that such problem occur only for matched/merged sample.
I have just pushed a patch in the development version for such problem.
You can download the patch here:
http://bazaar.launchpad.net/~mg5core2/mg5amcnlo/2.6.1/diff/282?context=3

Another way to fix your problem should be to set run_mode on 0 (single core) rather than 2 (multicore)

Cheers,

Olivier

> On 11 Sep 2017, at 19:10, Daniel Egana-Ugrinovic <email address hidden> wrote:
>
> ** Description changed:
>
> Hi,
>
> When showering with Pythia8 with just one core, showering fails. The
> error one gets is
>
> Pythia8 cross-section could not be retreived.
> Try turning parallelization off by setting the option nb_core to 1.
>
> The given message doesn't make sense, since I am actually running on a
> single core. Strangely, the problem only arises when working with only
> one core, with generate_events --nb_core=1. Using generate_events
> --nb_core=2 fixes the issue (unfortunately, when running on clusters,
> system administrators usually ask you to avoid multi-threaded code, so
> - in those cases using nb_core=2 would not help).
> + in those cases using nb_core=2 would not help). The problem may be that
> + the file unweighted_events.lhe is not being compressed into a gzip file
> + unweighted_events.lhe.gz. Since run_shower.sh seems to need a gzipped
> + lhe file, showering fails.
>
> Further information:
> Options are set in mg5 with
> set run_mode 2
> set nb_core 1
> save options
>
> Events are being run with madevent:
> generate_events runname --nb_core=1
> shower=PYTHIA8
>
> Technical details:
> Running madgraph 2.6.0
> Pythia8
> Python 2.7.13
> gcc (GCC) 6.2.0
>
> I attach the MG5 output file
>
> Thanks!
>
> Daniel
>
> --
> You received this bug notification because you are subscribed to
> MadGraph5_aMC@NLO.
> https://bugs.launchpad.net/bugs/1716425
>
> Title:
> Showering does not work when running with nb_core=1
>
> Status in MadGraph5_aMC@NLO:
> New
>
> Bug description:
> Hi,
>
> When showering with Pythia8 with just one core, showering fails. The
> error one gets is
>
> Pythia8 cross-section could not be retreived.
> Try turning parallelization off by setting the option nb_core to 1.
>
> The given message doesn't make sense, since I am actually running on a
> single core. Strangely, the problem only arises when working with only
> one core, with generate_events --nb_core=1. Using generate_events
> --nb_core=2 fixes the issue (unfortunately, when running on clusters,
> system administrators usually ask you to avoid multi-threaded code, so
> in those cases using nb_core=2 would not help). The problem may be
> that the file unweighted_events.lhe is not being compressed into a
> gzip file unweighted_events.lhe.gz. Since run_shower.sh seems to need
> a gzipped lhe file, showering fails.
>
> Further information:
> Options are set in mg5 with
> set run_mode 2
> set nb_core 1
> save options
>
> Events are being run with madevent:
> generate_events runname --nb_core=1
> shower=PYTHIA8
>
> Technical details:
> Running madgraph 2.6.0
> Pythia8
> Python 2.7.13
> gcc (GCC) 6.2.0
>
> ...

Read more...

Download full text (6.0 KiB)

Hi Olivier,

Thanks for the update! I tried with the patch, but the problem persists. I went ahead in the following order

- Copied the patch file which I named “singlecorepatch" into ./MG5_aMC_v2_6_0/madgraph/interface
- Applied the patch using patch madevent_interface.py singlecorepatch. patch Got the output "patching file madevent_interface.py”, and no errors.
- Generated the process from scratch, ./mg5 process_card.dat
- Ran the process exactly as I wrote down in my first post.

And then pythia8 fails with the error

PYTHIA Abort from Pythia::init: Les Houches Event File not found

This time I do not get the error "Try turning parallelization off by setting the option nb_core to 1”. Again, the file unweighted_events.lhe in the result folder appears uncompressed after the whole process is done. Is that normal? I thought it needed to be compressed for Pythia8 to find the file.

I attach the MG5 output and Pythia 8 log file.

Thanks!

Daniel

> On Sep 13, 2017, at 7:16 AM, Olivier Mattelaer <email address hidden> wrote:
>
> Hi Daniel,
>
>
> Ok looks like that such problem occur only for matched/merged sample.
> I have just pushed a patch in the development version for such problem.
> You can download the patch here:
> http://bazaar.launchpad.net/~mg5core2/mg5amcnlo/2.6.1/diff/282?context=3
>
> Another way to fix your problem should be to set run_mode on 0 (single
> core) rather than 2 (multicore)
>
> Cheers,
>
> Olivier
>
>> On 11 Sep 2017, at 19:10, Daniel Egana-Ugrinovic <email address hidden> wrote:
>>
>> ** Description changed:
>>
>> Hi,
>>
>> When showering with Pythia8 with just one core, showering fails. The
>> error one gets is
>>
>> Pythia8 cross-section could not be retreived.
>> Try turning parallelization off by setting the option nb_core to 1.
>>
>> The given message doesn't make sense, since I am actually running on a
>> single core. Strangely, the problem only arises when working with only
>> one core, with generate_events --nb_core=1. Using generate_events
>> --nb_core=2 fixes the issue (unfortunately, when running on clusters,
>> system administrators usually ask you to avoid multi-threaded code, so
>> - in those cases using nb_core=2 would not help).
>> + in those cases using nb_core=2 would not help). The problem may be that
>> + the file unweighted_events.lhe is not being compressed into a gzip file
>> + unweighted_events.lhe.gz. Since run_shower.sh seems to need a gzipped
>> + lhe file, showering fails.
>>
>> Further information:
>> Options are set in mg5 with
>> set run_mode 2
>> set nb_core 1
>> save options
>>
>> Events are being run with madevent:
>> generate_events runname --nb_core=1
>> shower=PYTHIA8
>>
>> Technical details:
>> Running madgraph 2.6.0
>> Pythia8
>> Python 2.7.13
>> gcc (GCC) 6.2.0
>>
>> I attach the MG5 output file
>>
>> Thanks!
>>
>> Daniel
>>
>> --
>> You received this bug notification because you are subscribed to
>> MadGraph5_aMC@NLO.
>> https://bugs.launchpad.net/bugs/1716425
>>
>> Title:
>> Showering does not work when running with nb_core=1
>>
>> Status in MadGraph5_aMC@NLO:
>> New
>>
>> Bug description:
>> Hi,
>>
>> When showering with Pythia...

Read more...

Hi,

Can you check it directly within 2.6.1?
You can get that beta version via the command : bzr branch lp:~mg5core2/mg5amcnlo/2.6.1

Cheers,

Olivier

To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers