Question in preprocessing own data about the r_distance parameter

1

Hi here, I just want to know for the following code is r_distance important?:
a2g = AtomsToGraphs(
max_neigh=50,
radius=6,
r_energy=True, # False for test data
r_forces=True, # False for test data
r_distances=False,
r_fixed=True,
)

Since when I read the original paper of dimenet and so on, they usually will utilize the edge_distance as a feature and transform it with some functions. In OC is these features important so we need to actually make r_distances = True when preprocess data?

2

Hi -

You do not need to specify it because the models will compute distance information as part of their forward pass. For example - fairchem/src/fairchem/core/models/dimenet_plus_plus.py at 1b4b62b4b71dfe755c94c60839f5d810e5d14a91 · FAIR-Chem/fairchem · GitHub.