[molpro-user] Way to ignore error and continue calculations without exit?

[Zhongqi JIN]

Hi Andy,

Thank you very much for your reply. You suggestion is useful. I have
tried that way of writing scripts. It did works. But it also brings
some problem. First, the time to deal with input and output files
increases dramatically,i.e. from one command to hundreds of command
for same amount of calculations. Second, due to the limitation numbers
of cpu allowed in using simultaneously on the server. This way of
writing scripts actually have low efficiency.

May I ask that your reply "This is not the case with some of the other
errors, and these can not therefore be skipped." is an Official Answer
or not? Because I think most of commercial software have some choice
to let you ignore/skip error and go on, in particular for this kind of
series of independent calculations. I also hope/highly doubt that
molpro should have one similarly function like that.

Best

Chung




On Wed, Jun 20, 2012 at 6:22 AM, Andy May <MayAJ1 at cardiff.ac.uk> wrote:
> Chung,
>
> In the case of convergence, the last iteration contains everything
> (integrals etc.) to continue on to perform a calculation with another
> method. This is not the case with some of the other errors, and these can
> not therefore be skipped.
>
> One solution would be to write a short bash script to generate an input file
> for each geometry. If you have a cluster then of course this is faster since
> you can run several inputs at one time.
>
> Best wishes,
>
> Andy
>
>
> On 19/06/12 14:37, Zhongqi JIN wrote:
>>
>> Hello all,
>>
>> I am using do loop to run a bunch of molpro calculations one after
>> another in one job, i.e. change the bond length/atom distance
>> repeatedly and calculate energy respectively. Sometimes, calculation
>> with one specific bond length failed and the whole job stop and exit.
>> I don't want them to stop when one of the calculations gives me an
>> error. I want molpro to ignore or bypass this error, stopping just
>> this calculation and continue to next bond length and finish the whole
>> job.
>>
>> For the convergence error, I can use nocheck card to ignore the error.
>>  But for other kind of error, such as “BASIS LINEARLY DEPENDENT OR
>> WRONG S” or “Insufficient overlap”. I have no idea how to do bypass
>> them.
>>
>> I have tried the STATUS, CLEAR and STATUS,IGNORE card, It did check
>> the succeeded calculations and give me OK status but when the
>> calculation failed it did not ignore the error and the job stop and
>> exit again.
>>
>> I have searched google and several forum but no luck. I really need
>> some help. Please help me!
>>
>> Thank you so much.
>>
>> Best Regards,
>>
>> Chung
>>
>> Attached my sample input file and Error message.
>>
>> ***,
>> memory,150,M
>> file,1,h3.int,new
>> file,2,h3.wfc,new
>>
>> gthresh,energy=1.0d-12
>> gthresh,twoint=1.0d-12
>>
>> geom={
>> x,z
>> noorient
>> zmat
>> x
>> x2 x 1.0
>> H3 x r x2 90
>> H1 x 0.5*rh2 x2 90 H3 theta
>> H2 x 0.5*rh2 x2 90 H1 180
>> endz
>> }
>>
>> basis={spdfg,h,aug-cc-pV5Z;c;}
>> rh2v=[2.801] bohr
>> thetav=[0.0] deg
>> rv=[2.9,2.7,2.5,2.3,2.1] bohr
>>
>> do j=1,#thetav
>>
>> theta=thetav(j)
>>
>> do i=1,#rh2v
>>
>> rh2=rh2v(i)
>>
>> do k=1,#rv
>> r=rv(k)
>>
>> {multi
>> maxiter,100;
>> start,atdens;
>>
>> config;
>> occ,3,0,0,0
>> closed,0,0,0,0
>> wf,3,1,1;state,2;
>> natorb,2143.2,ci,print=5}
>>
>> {ci,nocheck
>> maxiter,100;
>> state,1,2;
>> occ,3,0,0,0
>> closed,0,0,0,0
>> core,0,0,0,0
>> wf,3,1,1
>> natorb,3100.2,print=5
>> }
>> status,multi,ci,clear;
>> enddo; enddo; enddo
>> ---
>>
>>  ? Error
>>  ? Insufficient overlap
>>  ? The problem occurs in cihdia
>>
>>  GA ERROR fehler on processor   0
>> tmp =
>> /home/yunchung/pdir//share/apps/molpro/molpro2006.1/bin/molprop_2006_1_i8_x86_64_tcgmsg.exe.p
>>  Creating: host=compute-2d-0-7.local, user=yunchung,
>>
>>  file=/share/apps/molpro/molpro2006.1/bin/molprop_2006_1_i8_x86_64_tcgmsg.exe,
>> port=57208
>> _______________________________________________
>> Molpro-user mailing list
>> Molpro-user at molpro.net
>> http://www.molpro.net/mailman/listinfo/molpro-user
>>
>