Hi, thanks for sharing the code, which is a remarkable step over EGNN.
However, after looking deep into your code, I am confused about how SEGNN update the parctiple positions?
As the forward function shows, you only update the node features after feeding the graph into the model.
Therefore, how do you perform the N-body task to predict the final location of the particles?
Hi, thanks for sharing the code, which is a remarkable step over EGNN.
However, after looking deep into your code, I am confused about how SEGNN update the parctiple positions?
As the forward function shows, you only update the node features after feeding the graph into the model.
Therefore, how do you perform the N-body task to predict the final location of the particles?