[Feat] change ase.neighborlist to from matscipy.neighbours

[hexagonrose]

No description provided.

[YutackPark]

Thanks for the contribution. Some questions:

1. Why we use this for only unlabeled_** routine? It will make preprocessing faster if implemented to atoms_to_graph function.

2. It seems doing some hacky things if the given system is non-periodic. Is this tested for non-periodic systems?

[hexagonrose]

1. Oh, I missed the atoms_to_graph function, which also uses ase.neighborlist. In sevennet_calculator.py, it only uses unlabeled_atoms_to_graph, so I only fixed that part.

2. MACE also relies on those hacky things. I haven't yet tested it further, like MD simulations, but without these hacky things, it triggered a singular matrix error (molecules and isolated atom).

[YutackPark]

It's fine about atoms_to_graph, I just wondered if there exists some reason preventing implementaion. It is also absolutely fine to use hacky things, but if it is related to specific cases (non-pbc), it should be tested before merge. If my memory is correct, there already exists non-pbc calculator test. But double check is always a good idea. I'll check the first case (atoms_to_grpah). Thanks again for the contribution! 2024년 11월 7일 (목) 오후 10:31, hexagonrose ***@***.***>님이 작성:

…

1. Oh, I missed the *atoms_to_graph* function, which also uses *ase.neighborlist*. In *sevennet_calculator.py*, it only uses *unlabeled_atoms_to_graph*, so I only fixed that part. 2. MACE also relies on those hacky things. I haven't yet tested it further, like MD simulations, but without these hacky things, it triggered a singular matrix error (molecules and isolated atom). — Reply to this email directly, view it on GitHub <#124 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A2RQY2Z2ALY34WO55YZQEQDZ7NTTZAVCNFSM6AAAAABRKVP6VGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINRSGI2DSNZZGY> . You are receiving this because you commented.Message ID: ***@***.***>