Using ORCA via ASE with customized input file

[annulen]

After some experimentation I've found that the following command works:

geometric-optimize --ase-class=ase.calculators.orca.ORCA --ase-kwargs="{\"orcasimpleinput\":\"B3LYP D3BJ\", \"orcablocks\":`cat blocks.inp | jq -Rsa .`}"

where blocks.inp is my part of the input file including basis sets which obviously cannot be passed inline.

Is this the right way, or something more elegant is available? Anyway it would be great to have examples of such things in documentation. I guess a lot of people just don't realize they can use their software of choice with geomeTRIC, or give up trying to connect things.

[leeping]

It is great that you figured out how to do this. This isn't in the documentation and I did not know myself it was possible (though it is reasonable, given that ASE can interface to many different codes).

If you could contribute an example input file for a simple system that can run in a few minutes, I would include the example in the documentation.

[leeping]

With your permission I'd like to include the ORCA calculation you shared in the other thread as an example of using ORCA via ASE. If this is all right with you, could you please upload your input file? (I could try to copy it out of the output, but I am not sure if ORCA is modifying it.)

[annulen]

Here it is
blocks.inp.txt
But I think it's not small enough for quick example, and also as we've seen it's currently a nightmare to converge properly :)
Also it features basis set from 2023 paper which hasn't been thoroughly battle tested, and was actually designed for use with a different functional.
Perhaps it would be better to take some other example, like one with a few water molecules, and put only %pal and %scf blocks as a demonstration of what could be done. %pal is used to control number of parallel processes so it's the most important

BTW, it would be better if geomeTRIC could handle restart files of ORCA: orca.gbw and orca.ges. Right now consequential steps work fine as ORCA automatically restarts SCF in case of gradient jobs, however I've seen that geomeTRIC has a code to restore old orbitals in case of rejected step.

[leeping]

Thanks for the input file and the notes. If we access ORCA through ASE, then I hope we can use ASE to get the temporary file names too.