Re: [mdnalysis-discussion] Parallelize over a LAMMPS trajectory

[Oliver Beckstein]

Hi Thibault Flottat,

Can you isolate the part of your code that leads to the error?

What happens if you just run the lines 

import MDAnalysis as mda

data = 'structure.data'

dump = '1.26.Trajectory.lammpstrj'

u = mda.Universe(data, dump, lammps_coordinate_convention='unwrapped', format='LAMMPSDUMP’)

Also try just

u2 = mda.Universe(data)

Please paste the complete stack trace, because it contains useful information.

Do you have to explicitly indicate the atom style with the keyword atom_style as documented in https://docs.mdanalysis.org/stable/documentation_pages/topology/LAMMPSParser.html ?

Thanks,

Oliver

On Dec 4, 2021, at 4:53 AM, Thibaut <thibaut...@gmail.com> wrote:

Dear MDA users,

I am trying to parallelize my analysis over a LAMMPS trajectory. I am following the approach described here : https://userguide.mdanalysis.org/2.0.0-dev0/examples/analysis/custom_parallel_analysis.html, slicing trajectory in blocks.

Unfortunately, I have following errors, depending on how the universe was created, and seem to be related to the LAMMPS data reader :

ValueError: invalid literal for int() with base 10: '4280 7 1 C 15.499 13.5828 9.82833 -1 1 0 -51.49 80.5718 9.82833 -0.00251524 -0.00675299 -0.0014561\n'

ValueError: invalid literal for int() with base 10: '19708 828 4 H 18.6967 10.5701 12.2405 -1 1 0 -48.1262 77.393 12.2405 -0.00276927 0.0334039 -0.00127416\n'

ValueError: No coordinates following convention unscaled found in timestep

ValueError: No coordinates following convention unwrapped found in timestep

Here is the (very simple) code : https://codeshare.io/vwXpvl

How could I solve this ?

Thank you in advance.

--
You received this message because you are subscribed to the Google Groups "MDnalysis discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mdnalysis-discus...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/mdnalysis-discussion/0beae7e7-09a7-4927-be85-e5571d38dd1bn%40googlegroups.com.

--

Oliver Beckstein (he/his/him)

email: orbe...@mdanalysis.org
twitter: @orbeckst

MDAnalysis – a NumFOCUS fiscally sponsored project

https://www.mdanalysis.org/