The suggested torch-sim interface to relax crystal is to use ts.optimize. This function returns the state relaxed, whether or not the state has reached convergence. I believe this function should probably return a convergence boolean tensor, or the docs should encourage the user to run:
convergence = convergence_fn(final_state)
The suggested torch-sim interface to relax crystal is to use
ts.optimize. This function returns the state relaxed, whether or not the state has reached convergence. I believe this function should probably return aconvergenceboolean tensor, or the docs should encourage the user to run: