Problems in the web interface

[Johan Alwall]

There are a few things missing in the web interface ever since v. 1.4:
- Stop event generation button (this is very important, since there is
now no way to stop a gigantic generation that was started by mistake
or with mistakes in the cards).
- Remove banner. Also remove run seems bugged (the plots html page is
still there, but the plots are gone).
- Regenerate run from a removed run (i.e., from the banner of the
removed run). Instead, "run pythia" is there although it is
meaningless.
- Banner upload doesn't work, either to generate processes or to generate cards.

Additional problems on UIUC:
- Cluster status
- The SM calculators are not working
- The presentation of cross section and number of events after Pythia doesn't work

It also seems that the synchronization of user lists between clusters
is not working anymore. This can be quite annoying, since if a user
doesn't use exactly the same name (including uppercase/lowercase) when
they register on the two clusters, they won't get the same password.

[Olivier Mattelaer (olivier-mattelaer)]

Hi Johan,

- Stop event generation button (this is very important, since there is
now no way to stop a gigantic generation that was started by mistake
or with mistakes in the cards).

Ok I've finally a version which is working on madgraph.hep.uiuc.edu.
This is done in a very nice way (I suceed to send a ctrl-c to the python process and the python job takes care of stopping the jobs as if we stopped it in a shell). I will put the MG5 and WWW version under review and will be able to make the merge tommorow.

Will look at the rest later.

[Olivier Mattelaer (olivier-mattelaer)]

The synchronization of users is also in principle ready. I'll install it on monday (just to be sure with the school)

Cheers,

Olivier

[Olivier Mattelaer (olivier-mattelaer)]

I finally put in place the the synchronization of the users back between the cluster.

Cheers,

Olivier

[Olivier Mattelaer (olivier-mattelaer)]

Hi Johan,

> - Remove banner. Also remove run seems bugged (the plots html page is
> still there, but the plots are gone).
> - Regenerate run from a removed run (i.e., from the banner of the
> removed run). Instead, "run pythia" is there although it is
> meaningless.

This is fixed in 1.5.0. Could you try to start some test of the new functionality of 1.5.0? (the auto-update/the new options behavior/ …)

Also shouldn't we merge negative_weight inside 1.5.0?

Cheers,

Olivier

[Johan Alwall (johan-alwall)]

Hello Olivier,

Did you update branch 1.5.0 with the correct random seed treatment? It seems that the seed is not reused correctly in v. 1.5.0. Try to regenerate a process/run based on a banner with small statistics, and you will see that it is using a different seed.

Thanks,
Johan

[Olivier Mattelaer (olivier-mattelaer)]

Hi Johan,

Shouldn't this be fixed in 1.4.8?
This version is currently merged with 1.4.8 (I'm performing the merge with 1.4.8.2 right now)

Cheers,

Olivier

On Aug 17, 2012, at 3:54 AM, Johan Alwall <email address hidden> wrote:

> Hello Olivier,
>
> Did you update branch 1.5.0 with the correct random seed treatment? It
> seems that the seed is not reused correctly in v. 1.5.0. Try to
> regenerate a process/run based on a banner with small statistics, and
> you will see that it is using a different seed.
>
> Thanks,
> Johan
>
> --
> You received this bug notification because you are a member of MadTeam,
> which is subscribed to MadGraph5.
> https://bugs.launchpad.net/bugs/1028844
>
> Title:
> Problems in the web interface
>
> Status in The MadGraph Matrix Element Generator version 5:
> New
>
> Bug description:
> There are a few things missing in the web interface ever since v. 1.4:
> - Stop event generation button (this is very important, since there is
> now no way to stop a gigantic generation that was started by mistake
> or with mistakes in the cards).
> - Remove banner. Also remove run seems bugged (the plots html page is
> still there, but the plots are gone).
> - Regenerate run from a removed run (i.e., from the banner of the
> removed run). Instead, "run pythia" is there although it is
> meaningless.
> - Banner upload doesn't work, either to generate processes or to generate cards.
>
> Additional problems on UIUC:
> - Cluster status
> - The SM calculators are not working
> - The presentation of cross section and number of events after Pythia doesn't work
>
> It also seems that the synchronization of user lists between clusters
> is not working anymore. This can be quite annoying, since if a user
> doesn't use exactly the same name (including uppercase/lowercase) when
> they register on the two clusters, they won't get the same password.
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/madgraph5/+bug/1028844/+subscriptions

[Olivier Mattelaer (olivier-mattelaer)]

Hi Johan,

The problem is indeed already present in 1.4.8 (and probably much more older --maybe even since 1.0.0).

The point is that putting a value of the seed in the param_card IS NOT the same as putting zero and have it auto-assigned.

The reason is the following:
Each job read the banner in order to assign iseed.
If iseed is equal to zero each job read the file randinit in order to know what is the seed.
But the value inside randinit is modified at the refine level (+3 by refine) while the value in the run_card didn't change.

So multiple solutions:
1) We put in the run_card.inc the value "computed" at the beginning of the run since even with 0 we have the same behavior.
(But then we use the same seed for survey/refine1 and refine2)
2) We always use the value present in randinit (I would prefer not)
3) We put the seed as an input in input_app.txt/input_sg.txt (and we can take this opportunity to pass those fortran code in python)

What do you think?

Cheers,

Olivier

On Aug 17, 2012, at 2:55 PM, Olivier Mattelaer <email address hidden> wrote:

> Hi Johan,
>
> Shouldn't this be fixed in 1.4.8?
> This version is currently merged with 1.4.8 (I'm performing the merge with 1.4.8.2 right now)
>
> Cheers,
>
> Olivier
>
> On Aug 17, 2012, at 3:54 AM, Johan Alwall <email address hidden> wrote:
>
>> Hello Olivier,
>>
>> Did you update branch 1.5.0 with the correct random seed treatment? It
>> seems that the seed is not reused correctly in v. 1.5.0. Try to
>> regenerate a process/run based on a banner with small statistics, and
>> you will see that it is using a different seed.
>>
>> Thanks,
>> Johan
>>
>> --
>> You received this bug notification because you are a member of MadTeam,
>> which is subscribed to MadGraph5.
>> https://bugs.launchpad.net/bugs/1028844
>>
>> Title:
>> Problems in the web interface
>>
>> Status in The MadGraph Matrix Element Generator version 5:
>> New
>>
>> Bug description:
>> There are a few things missing in the web interface ever since v. 1.4:
>> - Stop event generation button (this is very important, since there is
>> now no way to stop a gigantic generation that was started by mistake
>> or with mistakes in the cards).
>> - Remove banner. Also remove run seems bugged (the plots html page is
>> still there, but the plots are gone).
>> - Regenerate run from a removed run (i.e., from the banner of the
>> removed run). Instead, "run pythia" is there although it is
>> meaningless.
>> - Banner upload doesn't work, either to generate processes or to generate cards.
>>
>> Additional problems on UIUC:
>> - Cluster status
>> - The SM calculators are not working
>> - The presentation of cross section and number of events after Pythia doesn't work
>>
>> It also seems that the synchronization of user lists between clusters
>> is not working anymore. This can be quite annoying, since if a user
>> doesn't use exactly the same name (including uppercase/lowercase) when
>> they register on the two clusters, they won't get the same password.
>>
>> To manage notifications about this bug go to:
>> https://bugs.launchpad.net/madgraph5/+bug/1028844/+subscriptions
>
> --
> You received this bug notification beca...

[Johan Alwall (johan-alwall)]

The seed issue should now be completely solved in 1.5.0.

All the best,
Johan