Is MadWeight tutorial still expected to work ?

Hi Jan,

MadWeight is for the moment in a debug only state.
We still support MadWeight but do not include any additional feature and this code is likely to pass as "not supported anymore".
at some point. Another code based on MadGraph: Momenta is actually the direct successor of that code.
(And I'm going to join the dev team of Momenta to assure his long term support)

But for the moment, I will happily fix any bug that you might face.

Based on previous experience, I claim that I am reasonably good at debugging stuff myself, and I will try to do this now. At the same time I would like to ask you two questions:

 - Is there a more recent version of this old tutorial ?

Actually a more recent tutorial (and more complete) is actually indicated on the page that you refer:
https://cp3.irmp.ucl.ac.be/projects/madgraph/raw-attachment/wiki/TOPMassMeasurmentExample/15_11_10_MadWeight_Tuto.pdf
the samples used are put on the cern indico:
https://indico.cern.ch/event/395374/contributions/939894/attachments/1184247/1716387/tutorial.tgz

 - Are you aware of any reason why this (MadWeight and/or the tutorial) should not work in the latest MG releases ? In other words: does it make sense for me to insist on trying to get this to run ?

No I'm not aware of any bug related to MW. It pass our test suite for the release in 2.6.1 and therefore should be working.
(one difference between the test suite and the online tutorial is that it does everything from the MG5 executable and do not call (directly) the
./bin/madweight.py executables.)
I think that's the reason which explains why we did not face the first bug that you mention.

For your second problem, can you check with the sample included in the CERN indico?
If this one goes trough, please open a bug report and attach your lhco file here, such that I can see why you do not have any events selected.

Cheers,

Olivier

PS: Sorry for the delay, last week I was in holliday, and I'm only part time working for the physics department.

On 16 Feb 2018, at 16:07, Jan Stark <<email address hidden><mailto:<email address hidden>>> wrote:

New question #664480 on MadGraph5_aMC@NLO:
https://answers.launchpad.net/mg5amcnlo/+question/664480

Dear experts,

I did not manage to run this old MadWeight tutorial:

 https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/TOPMassMeasurmentExample#no1

I used the latest version of MG: MG5_aMC_v2.6.1.tar

Even for the simplest example in the tutorial (top mass at the parton level), I did run into pretty basic problems:

 - python error when I run madweight.py (I think I fixed it)

 - no events selected in the check_events step

Based on previous experience, I claim that I am reasonably good at debugging stuff myself, and I will try to do this now. At the same time I would like to ask you two questions:

 - Is there a more recent version of this old tutorial ?

 - Are you aware of any reason why this (MadWeight and/or the tutorial) should not work in the latest MG releases ? In other words: does it make sense for me to insist on trying to get this to run ?

Thanks,
Jan

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

Dear Olivier,

Thank you for your clear and detailed answer. I will go on a long week-end myself tonight, so my next steps will not be immediate. But I will definitely report an update next week.

Cheers,
Jan

Dear Olivier,

Everything takes longer than expected. By I am now working with an intern on this project (assess performance of matrix element method for the measurement of the tri-linear Higgs self-coupling at the LHC). With the information that you have provided, we were able to run the (top mass) tutorial and we were able to adapt it to our purpose. Thank you again, we do appreciate your help !

Concerning the longer term:

  - Where can we learn more about the new Momenta tool that you mentioned ? Can we try it ??

  - We knew that CPU power is always an issue in these analyses. How about GPU support in MadWeight or Momenta ? Any project on this ?

Thanks,
Jan

Dear Olivier, Dear All,

OK, the new tool is called MoMEMta, not Momenta. [And of course, googling for "Momenta" had me drown in results.]

This is the webpage:

   https://momemta.github.io/

Unfortunately, I did not see any mention of GPUs on the MoMEMta page or in the paper that describes the tool that is used for integration in MoMEMta.

Cheers,
Jan

Hi Jan,

MadWeight does not support GPU. For Momenta, I do not know if this is not in the paper it is likely not supporting that type of architecture.

Cheers,

Olivier