Do not assume all pair_coeff come with a path

[pmrv]

At least for MTP potentials the format for the lammps potential is like

pair_style mlip mlip.ini
pair_coeff * *

where mlip.ini is the potential specific file. That means when
fix_paths tries to look a filepath to adjust in pair_coeff it won't find
anything and crash. This also means fix_paths might have to be applied
to pair_style also, but I haven't done that here. :/

I could not test further if there's issues with MTPs, because I'll need to custom build the lammps python first, so I am not sure if this change is the best way about it, but at least this quick fix got me somewhere calphy actually starts to try and run a calculation.

[srmnitc]

@pmrv This might have a bit more far reaching consequences. Similar style is used when fixing hybrid pair styles, for example here. I will check it with the pair style you provided above.

[srmnitc]

@pmrv Calphy uses pair hybrid/scaled method and applies the same potential twice. When this is done, would the commands look like this?

pair_style       hybrid/scaled 0.5 mlip mlip.ini 0.5 mlip mlip.ini
pair_coeff       * * mlip mlip.ini 1
pair_coeff       * * mlip mlip.ini 2

The 0.5s above are just the scale for each potential. I am not sure if this works, but could it be something that you can check? I feel specifying a file in the pair_style is not so common in LAMMPS, but I assume this would be too difficult to change.

[pmrv]

@pmrv This might have a bit more far reaching consequences. Similar style is used when fixing hybrid pair styles, for example here. I will check it with the pair style you provided above.

I had fear as much. :|

[pmrv]

@pmrv Calphy uses pair hybrid/scaled method and applies the same potential twice. When this is done, would the commands look like this?

pair_style       hybrid/scaled 0.5 mlip mlip.ini 0.5 mlip mlip.ini
pair_coeff       * * mlip mlip.ini 1
pair_coeff       * * mlip mlip.ini 2

The 0.5s above are just the scale for each potential. I am not sure if this works, but could it be something that you can check? I feel specifying a file in the pair_style is not so common in LAMMPS, but I assume this would be too difficult to change.

Sure, I'll give it a try manually on the cluster version today.