regarding PDF uncertainty while using LHAPDF for NLO models

Asked by Disha Bhatia on 2020-05-04

I was trying to run the tutorial which you had shown in the spring2020 DM t channel conference.
For PDF uncertainties, I am not getting the upper and lower bound errors as I am getting for scale variation.
I am attaching a screen shot of it below:
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
WARNING: Could not access LHAPDF to compute uncertainties for NNPDF30_nlo_as_0118
INFO:
   --------------------------------------------------------------
      Final results and run summary:
      Process p p > dm dm/ excluded [QCD]
      Run at p-p collider (6500.0 + 6500.0 GeV)
      Total cross section: 9.253e-01 +- 5.1e-03 pb
   --------------------------------------------------------------
      Scale variation (computed from histogram information):
          Dynamical_scale_choice -1 (envelope of 9 values):
              9.247e-01 pb  +1.8% -1.3%
      PDF variation (computed from histogram information):
          NNPDF30_nlo_as_0118 (101 members; combination method unknown):
              9.247e-01 pb
   --------------------------------------------------------------
%%%%%%%%%%%%%%%%%

Can you please help me in this regard?

Thank you,
Disha Bhatia

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Olivier Mattelaer
Solved:
2020-05-05
Last query:
2020-05-05
Last reply:
2020-05-05

Hi,

1) Did you use the internal PDF or did you use lhapdf?
2) Do you have lhapdf installed?

Cheers,

Olivier

Disha Bhatia (dishabhatia1989) said : #2

Hi,

I had installed lhapdf6 using the install command in the ./bin/mg5 terminal.

But when I ran my program, it complained MadGraph5Error : Could not download NNPDF30_nlo_as_0118 into /home/disha/software_required/MG5_aMC_v2_6_7/HEPTools/lhapdf6//share/LHAPDF. Please try to install it manually.

Then I downloaded NNPDF30_nlo_as_0118 manually and had placed it in lhapdf6/share/LHAPDF folder
inside Madgraph.

Now the code is running fine, but is not giving PDF uncertainites.
I am using MG5_aMC_v2_6_7 version.

On which value is the parameter "pdlabel" in the run_card?
@NLO, you should have it to lhapdf to be able to compute the associated error.

Cheers,

Olivier

Disha Bhatia (dishabhatia1989) said : #4

Yes, I had declared it as lhapdf only. Here is a screenshot of that.

***********************************************************************
# PDF choice: this automatically fixes also alpha_s(MZ) and its evol. *
#***********************************************************************
 lhapdf = pdlabel ! PDF set
 260000 = lhaid ! If pdlabel=lhapdf, this is the lhapdf number. Only
              ! numbers for central PDF sets are allowed. Can be a list;
              ! PDF sets beyond the first are included via reweighting.

It asks me to manually install NNPDFset and with that I don't get uncertainties.

Thanks,
Disha

My advice, would then to move to 2.7.2 and re-install lhapdf and retry.
moving to the latest version of MG5aMC will allow to install lhapdf6.2 instead of lhapdf6.1 (which has a lot of issue with the dataset...)

Cheers,

Olivier

Disha Bhatia (dishabhatia1989) said : #6

Ok I will do that, if I my doubt persist then I will write back to you.

Thank you,
Disha

Disha Bhatia (dishabhatia1989) said : #7

I tried with 7.2 version, it didnot ask me to manually download the pdf's, it did it on its own,
however the error still persists.

WARNING: Could not access LHAPDF to compute uncertainties for NNPDF30_nlo_as_0118

Hi,

I have tried to reproduce your issue and I do not reproduce it.
Looks like your have an issue with your lhapdf configuration.
In order to have more details,
you can change line 4659 of madgraph/interface/amcnlo_run_interface.py
and/or bin/internal/amcnlo_run_interface.py (depending of the method that you use to run the code)

you should have:
                    except:
                        logger.warning("Could not access LHAPDF to compute uncertainties\
 for %s" % pdfsetname)

and you can replace those by
                    except:
                        raise
                        logger.warning("Could not access LHAPDF to compute uncertainties\
 for %s" % pdfsetname)

This should give more information on why lhapdf fails to provide the information about which type of uncertainty need to be used.
This being said the lhef file should have all the information needed for you to compute such error (by calling the associate lhapdf routine).

Cheers,

Olivier

Disha Bhatia (dishabhatia1989) said : #9

Thank you so much for the suggestion.

I am now getting this warning:

WARNING: Failed to access python version of LHAPDF: cannot compute PDF uncertainty from the weights in the events. The weights in the LHE event files will still cover all PDF set members, but there will be no PDF uncertainty printed in the run summary.
 If the python interface to LHAPDF is available on your system, try adding its location to the PYTHONPATH environment variable and theLHAPDF library location to LD_LIBRARY_PATH (linux) or DYLD_LIBRARY_PATH (mac os x).

Following this, I tried to set the environment in .bashrc and did source .bashrc but probably I need to declare it somewhere else.
Because it didn't work.

not sure that this is the correct error, maybe then change those line to

                    except Exception as error:
                        print error
                        sys.exit(0)
                        logger.warning("Could not access LHAPDF to compute uncertainties\
 for %s" % pdfsetname)

Cheers,

Olivier

Yes you were right, earlier it was not giving full information.

I tried this and now it doesn't run full and give this message

 Results after grid setup:
      Total cross section: 9.400e-01 +- 8.2e-03 pb

INFO: Refining results, step 1
INFO: Idle: 0, Running: 2, Completed: 0 [ current time: 21h35 ]
INFO: Idle: 0, Running: 1, Completed: 1 [ 1m 18s ]
INFO: Idle: 0, Running: 0, Completed: 2 [ 1m 21s ]
'module' object has no attribute 'getPDFSet'

Thank you,
Disha

Hi,

My guess is that you have a module call lhapdf somewhere in your PYTHONPATH and that however that one is not a lhapdf code (or a very old one).

Cheers,

Olivier

> On 5 May 2020, at 14:58, Disha Bhatia <email address hidden> wrote:
>
> Question #690431 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/690431
>
> Status: Answered => Open
>
> Disha Bhatia is still having a problem:
> Yes you were right, earlier it was not giving full information.
>
> I tried this and now it doesn't run full and give this message
>
> Results after grid setup:
> Total cross section: 9.400e-01 +- 8.2e-03 pb
>
> INFO: Refining results, step 1
> INFO: Idle: 0, Running: 2, Completed: 0 [ current time: 21h35 ]
> INFO: Idle: 0, Running: 1, Completed: 1 [ 1m 18s ]
> INFO: Idle: 0, Running: 0, Completed: 2 [ 1m 21s ]
> 'module' object has no attribute 'getPDFSet'
>
> Thank you,
> Disha
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Yes you were right, I finally managed to fix it.
Thank you so much for the help and all suggestions.

Thanks,
Disha

Thanks Olivier Mattelaer, that solved my question.