[ase-users] problem with dftd3 when running calculator in parallel

[K. Doblhoff-Dier]

Dear Eric,

Jens Jørgen already fixed it on master! (see mailing list 03/04/2019)

Thanks anyway,

Best regards,

Katharina


On 06/05/2019 16:39, Hermes, Eric wrote:
>
> Apologies for the late response, I don't check the mailing list as 
> often as I should.
>
>
> This is something that I can and will fix. Please provide a minimal 
> working example (the smallest full script you can make that reproduces 
> this issue), and I will debug the issue further.
>
>
> Eric Hermes
>
> ------------------------------------------------------------------------
> *From:* ase-users-bounces at listserv.fysik.dtu.dk 
> <ase-users-bounces at listserv.fysik.dtu.dk> on behalf of K. 
> Doblhoff-Dier via ase-users <ase-users at listserv.fysik.dtu.dk>
> *Sent:* Wednesday, April 3, 2019 3:32 AM
> *To:* ase-users at listserv.fysik.dtu.dk
> *Subject:* [EXTERNAL] [ase-users] problem with dftd3 when running 
> calculator in parallel
>
> Dear ase community,
>
> I tried to do a dftD3 calculation on top a gpaw calculation. As far as 
> I have understood, dftD3 does not run in parallel, but compared to the 
> DFT calculation, I guess it will not be a bottleneck. However, while 
> everything works fine if I run the gpaw calculation in serial, it does 
> not, if I run my script with
>
> mpirun  gpaw-python gpaw_runscript.py
>
> with a gpaw_runscript.py looking as follows
>
> dft=GPAW(..., parallel=dict(...))
> calc=DFTD#(dft=dft)
> atoms.set_calculator(calc)
> atoms.get_potential_energy()
>
> The error I get is the following:
>
> rank=00 L09:   File "gpaw_runscript.py", line 120, in <module>
> rank=00 L10:     slab.get_potential_energy()
> rank=00 L11:   File ".../python2.7/site-packages/ase/atoms.py", line 
> 664, in get_potential_energy
> rank=00 L12:     energy = self._calc.get_potential_energy(self)
> rank=00 L13:   File 
> ".../python2.7/site-packages/ase/calculators/calculator.py", line 505, 
> in get_potential_energy
> rank=00 L14:     energy = self.get_property('energy', atoms)
> rank=00 L15:   File 
> ".../python2.7/site-packages/ase/calculators/dftd3.py", line 395, in 
> get_property
> rank=00 L16:     allow_calculation)
> rank=00 L17:   File 
> ".../python2.7/site-packages/ase/calculators/calculator.py", line 552, 
> in get_property
> rank=00 L18:     self.calculate(atoms, [name], system_changes)
> rank=00 L19:   File 
> ".../python2.7/site-packages/ase/calculators/dftd3.py", line 238, in 
> calculate
> rank=00 L20:     errorcode = self.comm.broadcast(errorcode, 0)
> rank=00 L21: TypeError: Not a proper NumPy array for MPI communication.
> GPAW CLEANUP (node 0): <type 'exceptions.TypeError'> occurred.  
> Calling MPI_Abort!
>
> Is it possible to have gpaw run in parallel and put dftD3 on top?
>
> Thank you and best regards,
>
> Katharina Doblhoff-Dier
> Leiden University
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.fysik.dtu.dk/pipermail/ase-users/attachments/20190506/7f96dd1f/attachment-0001.html>