qcscine / qcmaquis

Release-only repository for SCINE QCMaquis, the DMRG software from the Reiher group.
BSD 3-Clause "New" or "Revised" License
32 stars 15 forks source link

Problem setting up AutoCAS #9

Open westrangeloops opened 2 years ago

westrangeloops commented 2 years ago

Hi, I'm a student from Colombia. I would like to use AutoCas for research purpose, currently have OpenMolcas - Qcmaquis (make install and set the PATH=$PATH) interface which work well, pymolcas qcmaquis verify only fail in 007 test and the interface ethane calculation work fine too (pymolcas ethane.inp)

I compile it in my work computer (I12700k 32gb ram and Ubuntu distro, I would expand the ram later), but when I'm trying to use AutoCas for the butadiene or ethyelene examples it fails with the message: "Apparently your initial orbital generation did not run..." so I enter my project folder and found that the HF don't to nothing, if run pymolcas initial_orbital.inp calculation works well and the files Initial_orbitals.n (n = GssOrb, ScfOrb, guessorb.h5 and molden) are generated in the project directory, but after I manually run the inp file AutoCas still don read the orbitals.

I don't really understand which orbital files are needed for a correct AutoCAS read, or if I'm not correctly setting my run.env

My current OpenMolcas src is:

! molcasrc configuration file written by /opt/OpenMolcas/pymolcas -setup

MOLCAS_PROJECT=NAME MOLCAS_MEM=28000 MOLCAS_MAXITER=100 MOLCAS_NEW_DEFAULTS=NO MOLCAS_TRAP=YES MOLCAS_NEW_WORKDIR=YES MOLCAS_KEEP_WORKDIR=NO MOLCAS_SAVE=INCR MOLCAS_MOLDEN=YES MOLCAS_PRINT=NORMAL MOLCAS_ECHO_INPUT=YES MOLCAS_COLOR=YES MOLCAS_REDUCE_PRT=YES

Maybe I'm setting bad my run.env which currently run.env is

!/bin/bash

! set MOLCAS3

export myMOLCAS=/opt/OpenMolcas

! source $myMOLCAS/qcmaquis/bin/qcmaquis.sh This file is not created in my installation

export QCMaquis_CPUS=12 input=$1 project={{project_name}} scratch=/scratch/${USER}

CurrDir=$PWD export MOLCAS=$myMOLCAS export MOLCAS_PROJECT=$project export MOLCAS_MEM=28000 export MOLCAS_WORKDIR=${scratch} pymolcas -b 1 -f $CurrDir/$input

Really appreciate your help, Regards Camilo Diaz.

mmoerchen commented 2 years ago

Hi Camilo,

I'm one of the AutoCAS developers. There is a bug in your run.env script. In the last line, you call pymolcas, without specifying the location of that script. Also your myMOLCAS variable should point to the build directory of OpenMolcas. The reason for that behavior is, that internally AutoCAS starts pymolcas in a new shell, so that all environment variables have to be set in the environment file. An example setup for run.env, could be:

#!/bin/bash

export myMOLCAS=/path/to/OpenMolcas/build
export QCMaquis_CPUS=6
# export other MOLCAS environment variables here

input=$1
project={{project_name}}
scratch=/scratch/${USER}

CurrDir=$PWD
export MOLCAS=$myMOLCAS
export MOLCAS_PROJECT=$project
export MOLCAS_MEM=10000
export MOLCAS_WORKDIR=${scratch}
$myMOLCAS/pymolcas -b 1 -f $CurrDir/$input

Best, Max

westrangeloops commented 2 years ago

Hi Max, Thanks for your help, now my AutoCAS is working. I found some minor issues with the templates, for example the GETAWAY and SEWARD block are not wrote when I set up the DMRG calculation and Final calculation, but this is not an issue.

However, my AutoCAS don't reproduce the same graphics as the examples, the area of the red circles is displayed correctly, but no lines are showed (red, black, green or gray dotted), so I was wondering that maybe I'm not correctly setting the correct options in the input file. I hope it is not a presumptuousness to ask you for some examples of a correct input file for the DMRG.

The other issue I have is that sometimes the active space for the final calculation is not selected according to the threshold or automatic selection, instead all orbitals are selected as active space, it happens particularly with only one system.

Best, Camilo.

mmoerchen commented 1 year ago

Hi Camilo,

sorry for the late reply, maybe you already fixed your problems.

The input for DMRG and FINAL uses the same orbitals specified in INITIAL, so GETAWAY and SEWARD are not required for these calculations. I never had issues with the display of mutual information, but this should be only a graphical but. Is the suggested CAS by AutoCAS reasonable for this system to select all orbitals?

By the way, we released a CLI version of AutoCAS, in case you want to check it out.

Best, Max