ok,
I think you problem was that you had not activated Q.SENSOR_RESPONSE.BACKEND_DO
= true. So it should work now.
The new error is something I have no idea about unfortunately, my only
suggestion if to make sure that you have the newest version of arts 2.2 and
atmlab 2.2. installed (and compiled).
You can also try and run atmlab in debug (write atmlab('DEBUG',1)) mode such
that your temporary arts files are saved. The go to the cfile.arts and see if
you see something strange in the call of jacobianClose.
Ole Martin
________________________________________
From: Fran Navas [[email protected]]
Sent: Tuesday, July 21, 2015 3:23 PM
To: Ole Martin Christensen
Cc: [email protected]; ARTS Users List
Subject: Re: [arts-users] arts 2.2
Dear Ole,
as you suggested I have changed the frequency grid (Q._GRID) to a larger
number and I have confirmed that also the jacobian matrix changed the
dimensions. However, after activating the backend
(Q.SENSOR_RESPONSE.BACKEND_DO = true) I did not achieve to solve the
problem and the next error message came out:
L2 = qpack2( Q, O, Y);
terminate called after throwing an instance of 'std::runtime_error'
what(): Run-time error in method: jacobianClose
Method jacobianClose needs input variable: sensor_response
/opt/arts/arts2.2/build/src/arts -r00 -o
/tmp/atmlab-tempera-tpa6df5bc9_d962_48a4_9b91_5f0713f5df5e
/tmp/atmlab-tempera-tpa6df5bc9_d962_48a4_9b91_5f0713f5df5e/cfile.arts:
Aborted
Regarding the f_backend8f.xml and f_backend.xml the number of elements
are 8 and 12 respectively. I attach again the function that set the
q_structure and also the backend files just in case that you want to
check is there is something wrong.
Once again, thank you very much for your help!
Best regards,
Fran
On 07/20/2015 06:46 PM, Ole Martin Christensen wrote:
> I notice that your frequency grid has 14 elements, so another explanation
> could be that the backend channels have not been activated in ARTS. You can
> check this by changing your frequency grid (Q.F_GRID) to a larger number of
> frequencies and see if the size of the jacobian matrix (J) you put into oem
> changes.
>
> if this is the case I think you might have to set
> Q.SENSOR_RESPONSE.BACKEND_DO = true
> to active the backend.
>
> another possiblity is that you have to many elements in in your
> f_backend8f.xml and f_backend.xml file? from your error I would almost guess
> that there are 14 elements in these files.
>
> I hope this helps, or at least diagnoses the problem.
>
> Ole Martin
>
>
> ________________________________________
> From: Fran Navas [[email protected]]
> Sent: Monday, July 20, 2015 6:09 PM
> To: Ole Martin Christensen
> Subject: Re: [arts-users] arts 2.2
>
> Dear Ole,
> I attach the functions for the setup of the Q and the Y structures and
> also the code of the inversion. I hope that it is clear for you and if
> you have any question just let me know.
>
> After your comment I am thinking that the problem could come from the
> Y.F variable. With ARTS version 2.0 this variable was empty in my
> setting and it worked, but now with the last version of ARTS the next
> error message appeared: ("Row size of Y(1).TNOISE must be 1 or equal to
> the length of Y.F"). For this reason I set Y.F with the frequencies of
> our radiometer and with the same dimension than Y.TNOISE, it means that
> the frequencies are repeated following the block measurements (12
> frequencies*9 elevations angles = 108 values). After this change I did
> not get anymore the error message but the one that I explained in my
> previous email.
>
> Well, I hope that you can help me with this problem and I finally can
> run the retrieval in ARTS 2.2 :) Thank you so much in advance for your help.
>
> Best regards,
> Fran
>
>
>
>
> On 07/19/2015 02:58 PM, Ole Martin Christensen wrote:
>> Hi, Fran
>>
>> (I'm adding the qpack mailing list here as well)
>>
>> could you attach your Q file and perhaps the script you use to execute the
>> retrievals? I assume this problem could be due to some mismatch between the
>> number of channels specified in you Y.F (qpack Y struckture) and
>> Q.SENSOR_RESPONSE.
>>
>> Ole Martin
>> ________________________________________
>> From: [email protected] [[email protected]]
>> on behalf of Fran Navas [[email protected]]
>> Sent: Wednesday, July 15, 2015 10:15 AM
>> To: ARTS Users List
>> Subject: [arts-users] arts 2.2
>>
>> Dear Patrick, Jana and all,
>> I am having some problems running our tropospheric temperature
>> retrievals (from filter bank measurements) in ARTS 2.2. These retrievals
>> were running in arts 2.0 until now. I have adapted the scripts in order
>> to run it in the last version and now everything look fine until the
>> different iterations start in the retrieval.
>>
>> Below you can find the error message. I do not why but the dimensions of
>> the 'yf' that is calculated in the function: [R,yf,J] = calc_y_and_j(
>> comfun, Q, R, O, xnorm, x, iter ); are different to the ones from 'y'.
>> In my retrievals (y: [108*1]; yf: [126*1]). This difference in the
>> dimensions is the cause of the error. I guess that something is wrong in
>> the Q-structure but I don't find what can be. I would appreciate a lot
>> if someone could help me.
>>
>> Thank you very much in advance!!!
>> Fran Navas
>>
>>
>> Error in ==> oem>calc_cost at 945
>> dd = y - yf;
>>
>> Error in ==> oem>calc_cost_for_X at 956
>> [cost,cost_y,cost_x] = calc_cost( y, yf, Seinv, xa, x, Sxinv, O,
>> xnorm );
>>
>> Error in ==> oem at 558
>> X = calc_cost_for_X( y, yf, Seinv, xa, x, Sxinv, O, xnorm, X,
>> iter );
>>
>> Error in ==> qpack2 at 116
>> [X,R] = oem( O, Qm, R, @arts_oem, Sx, Se, Sxinv, [], xa, Y(m).Y );
>>
>>
>>
>>
>>
>> --
>> *****************************************************************************
>> Francisco Navas Guzmán, PhD
>>
>> Microwave Physics Group
>> Division of Atmospheric radiometry
>> Institute of Applied Physics
>> University of Bern
>>
>> Address: | http://www.iap.unibe.ch
>> Sidlerstrasse 5 | Ph: +41-31 631 50 46
>> CH-3012 Bern (Switzerland) | Fax: +41-31 631 37 65
>> *****************************************************************************
>>
>> _______________________________________________
>> arts-users mailing list
>> [email protected]
>> https://www.sat.ltu.se/mailman/listinfo/arts-users
>
> --
> *****************************************************************************
> Francisco Navas Guzmán, PhD
>
> Microwave Physics Group
> Division of Atmospheric radiometry
> Institute of Applied Physics
> University of Bern
>
> Address: | http://www.iap.unibe.ch
> Sidlerstrasse 5 | Ph: +41-31 631 50 46
> CH-3012 Bern (Switzerland) | Fax: +41-31 631 37 65
> *****************************************************************************
>
--
*****************************************************************************
Francisco Navas Guzmán, PhD
Microwave Physics Group
Division of Atmospheric radiometry
Institute of Applied Physics
University of Bern
Address: | http://www.iap.unibe.ch
Sidlerstrasse 5 | Ph: +41-31 631 50 46
CH-3012 Bern (Switzerland) | Fax: +41-31 631 37 65
*****************************************************************************
_______________________________________________
qpack mailing list
[email protected]
https://www.sat.ltu.se/mailman/listinfo/qpack
