OCPCalculator a2g does not use edge nor distance information

1

Hello,
I have a quick question regarding the necessary information to run inference using the OCPCalculator.

I noticed that in the OCPCalculator class, the AtomsToGraphs converter has r_distances and r_edges set to False. As I understand it, the distance and edge formation is necessary for the models in the backend to run predictions.

So why is it possible to use the OCPCalculator with a model in the backend that requires that information, without storing that information in the AtomsToGraphs conversion process?

Thank you.

2

Hi,

It is not necessary to compute that information with the AtomsToGraphs method because the model computes the graph on the fly (otf_graph flag). fairchem/src/fairchem/core/models/base.py at 6efc99b8164887ce7d699ee6113d9168b4a3e9cc · FAIR-Chem/fairchem · GitHub

Best,
Brandon