Skip to content

Conversation

@ErjieWu
Copy link
Collaborator

@ErjieWu ErjieWu commented Apr 2, 2025

Linked Issue

Fix #6075

What's changed?

  • Add ticks in update_dmr() and double check the place it is used in DeePKS to avoid repeating calculations.
  • Use slice() method in torch::Tensor to simplify the code.
  • Combine the repeating loop for stress and force calculations in DeePKS.

@ErjieWu ErjieWu added The Absolute Zero Reduce the "entropy" of the code to 0 Machine Learning Issues related to the DeePKS labels Apr 2, 2025
@ErjieWu
Copy link
Collaborator Author

ErjieWu commented Apr 3, 2025

Note: The integrate test 204_NO_KP_NC (non-collinear Fe Anti-ferromagnetic and lcao basis calculation) results are different in runs and I'm not sure whether this is normal.

@mohanchen mohanchen merged commit ab2e27a into deepmodeling:develop Apr 3, 2025
14 checks passed
@ErjieWu ErjieWu deleted the perf branch April 3, 2025 09:09
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Machine Learning Issues related to the DeePKS The Absolute Zero Reduce the "entropy" of the code to 0

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Irreasonable time consumption in cal_pdm() in DeePKS.

2 participants