Alexander Goscinski

I don't know how to get the central atom tag from a ghost atom in lammps, because we use the neighbour tag of the nonghost atom when iterating through the...

Seems to work now, at least no segmentation fault for the input I used (methane_pbc.lammps https://github.com/agoscinski/lammps/blob/a5e34127bd55dfedebeae510fac8b88511837067/examples/rascal/methane_pbc.lammps). The neighbourlists using StructureManagerLammps and Center have same number of pairs with same tags,...

I just checked for two H atoms, the representation for the two atoms are not even the same with lammps. I will see if can get some useful information from...

I found the bug for the soap representation of two hydrogens (without pbc) being different: StructureManagerLammps is considered as half neighbour list in its traits. It is also written in...

The MPI support seems to work now. There are small errors when comparing the MPI runs for different number of tasks, this however seems lammps related, since I could also...

We should experiment as well how the array api works with our selection methods. FPS is probably a good candidate because we do not use very complicated mathematical operations there....

Hello, was able to pin point it to this line https://github.com/cosmo-epfl/librascal/blob/db2e2445d34c196c94731249061740123f9fbc28/src/rascal/representations/calculator_spherical_invariants.hh#L1373 Later when the gradients are resized using the `key_list_grad` the memory starts to differ between normalize True and False...

see updates PR summary at top

> > Current issues include benchmark tutorial (c++ based, needs to be compiled, #202 ), which cannot be housed in docs/source/tutorials for the compiler to locate it. > > This...

Could we add this small change https://github.com/cosmo-epfl/librascal/pull/367/files#diff-ee5297016f816d41032a665df5ca536e26f6eabae71f698a1f45036ba12342ca for better compatibility with the cpp side. So the model json can be also loaded very simply from the cpp side.