Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Refactor: Remove global dependence of descriptor, orbital_precalc, v_delta_precalc in DeePKS. #5808

Closed
wants to merge 11 commits into from

Conversation

ErjieWu
Copy link
Collaborator

@ErjieWu ErjieWu commented Jan 4, 2025

What's changed?

  • Remove d_tensor(descriptor) from LCAO_deepks and change functions related to descriptor in DeePKS.
  • Split the function of force and stress calculations cal_gdmx() in DeePKS into two separate functions cal_gdmx() and cal_gdmepsl(), rearrange the calculation related to DeePKS in FORCE_STRESS.cpp to make it same as other parts for force&stress calculations.
  • Remove the global dependence of orbital_precalc and v_delta_precalc in DeePKS and put their functions together in deepks_orbpre.cpp and deepks_vdpre.cpp.
  • Use accessor for tensor to accelerate the access/modify of torch::Tensor variables.
  • Remove LCAO_deepks_mpi.cpp since the function allsum_deepks() is no longer used now.

@ErjieWu
Copy link
Collaborator Author

ErjieWu commented Jan 4, 2025

I found that in the old store structure, the last two index m1 and m2 are stored by access [m1 * nm + m2], here nm is the total number of this l. However, in new structure, it will change to [m1][m2], and the size is nm_max*nm_max, where nm_max correspond to l_max. So in the checking (printing), the position of zeros will be in different positions, which cause the failure of UT. I will change the result.ref file to make it consistent with the new data structure in a new pr.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant