search

globus-labs / faster-molecular-hessians

Reduce the amount of computation needed for Hessians using machine learning
MIT License
0 stars 1 forks source link

Use Dscribe's MBTR derivatives to accelerate hessian computation #10

Open WardLT opened 1 year ago

WardLT commented 1 year ago

We use numerical derivatives in the MBTR to get forces, which is slow. You can use the ones from MBTR, and there's an example showing how to do so.

WardLT commented 10 months ago

We need derivatives of energies wrt descriptors to make this work.