[molpro-user] Assistance saving CI vectors for later use
Alexander.Mitrushchenkov at u-pem.fr
Mon Nov 30 14:39:35 CET 2015
You were right, and multi in the second input does not help to properly
set up arrays. To avoid any problem like that, please rerun (BOTH) jobs
putting somewhere in the beggining of the input "gparam,ibank=1" (for
BOTH inputs). Then you do not need multi in second file at all. Please
note that the binary files will most probably be not fully compatible
with default (i.e. ibank=2) runs. In principle, ibank=1 may affect the
performance for very big calculations, with large size matrix
multiplications, vector machines etc etc, but probably it is not really
that important, especially for modern computers.
Le 30/11/2015 11:59, Laura McKemmish a écrit :
> Hi Alexander,
> Thanks for the advice here. I was wondering if you could help me
> correct the following input files (which are currently not working).
> As far as I know, they are both using the same Molpro environment
> (number of processors, same version etc); however, there is a
> possibility the node type is different. I think, however, it is
> probably a more basic mistake in my input.
> FIRST FILE:
> SECOND FILE:
> On Sat, Nov 28, 2015 at 9:39 AM, Alexander Mitrushchenkov
> <Alexander.Mitrushchenkov at u-pem.fr
> <mailto:Alexander.Mitrushchenkov at u-pem.fr>> wrote:
> To save CI vectors, the only way is to use binary files. However,
> the molpro binary files are not supposed to be transferable. These
> are very likely to be incompatible between different molrpo
> versions, and even the same version compiled with different
> parameters. Also, if I well remember, there can be incompatibily
> between serial/parallel runs. So if you use it, ensure that molpro
> environment is the same.
> With CI vectors there might be one additional problem related to
> internal molpro architecture (in particular ntb and related
> arrays). When restarting CI vectors, I advise to perform some
> MULTI calculation, even if you do not need it, before using
> restart CI records. This should properly set up internal arrays.
> Le 26/11/2015 20:58, Laura McKemmish a écrit :
> I am looking to produce MRCI wavefunctions from separate
> calculations for different electronic states, then use them in
> a latter combination (e.g. to find transition moments using
> the biorth command etc).
> Looking at the manual and previous questions, there seems to
> may be two ways I can do this:
> (1) Use the Molpro binary files or
> (2) Use the Matrop to print out the MRCI vectors, then read
> them in later (not sure if this file size is prohibitive).
> I have made both of these options work sometimes for saving
> CASSCF orbitals, and the first work sporadically for saving
> MRCI wave functions. I am happy with the second option
> (human-readable file) for the CASSCF orbitals as it seems to
> be reliable and is quite portable. I am running into a lot of
> difficulties storing binary files in one job, then reusing
> them in another job. It sometimes works (when I do things in
> one order) but not other times (when I do something slightly
> different, though I am not sure what has made the difference).
> If I transport the same *.wfu file between multiple jobs, when
> do my records get overwritten?
> Any advice on the above would be very much appreciated. Also,
> any way to get information about what records are stored in a
> particular binary file would be great to know. Should I be
> storing orbitals and/or CI wave functions in certain records
> (e.g. 3001.2?)
> Thanks very much for any assistance/ advice with this.
> Kind Regards
> Molpro-user mailing list
> Molpro-user at molpro.net <mailto:Molpro-user at molpro.net>
> Dr. Alexander Mitrushchenkov, IGR
> Laboratoire de Modélisation et Simulation Multi Echelle
> UMR 8208 CNRS
> Université Paris-Est Marne-la-Vallée
> 5 Bd Descartes
> 77454 Marne la Vallée, Cedex 2, France
> Phone: +33(0)160957316 <tel:%2B33%280%29160957316>
> Fax: +33(0)160957320 <tel:%2B33%280%29160957320>
> e-mail: Alexander.Mitrushchenkov at u-pem.fr
> <mailto:Alexander.Mitrushchenkov at u-pem.fr>
Dr. Alexander Mitrushchenkov, IGR
Laboratoire de Modélisation et Simulation Multi Echelle
UMR 8208 CNRS
Université Paris-Est Marne-la-Vallée
5 Bd Descartes
77454 Marne la Vallée, Cedex 2, France
e-mail: Alexander.Mitrushchenkov at u-pem.fr
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Molpro-user