Skip to content

Conversation

@xuan112358
Copy link
Collaborator

Linked Issue

Fix #5113

What's changed?

  • Restore deepks unit test in gamma-only case
  • In deepks operator, multi-k case also has full size of Hamiltonian of DeePKS now.
    In the past time, multi-k keep same size of Hamiltonian as nonlocal operator. This is not accurate.
    This modification will increase deepks calculation time for multi-k case. To reduce time, also modify some STRUs in integrate test s of deepks.

xuan112358 and others added 27 commits October 26, 2024 13:44
…D in deepks unit test instead of GlobalC::GridD
…or complete atom_pairs will cause time increase in multi-k cases.
@mohanchen mohanchen added the Machine Learning Issues related to the DeePKS label Dec 31, 2024
@mohanchen mohanchen added Refactor Refactor ABACUS codes Compile & CICD & Docs & Dependencies Issues related to compiling ABACUS labels Jan 4, 2025
@mohanchen mohanchen merged commit d061719 into deepmodeling:develop Jan 4, 2025
14 checks passed
Fisherd99 pushed a commit to Fisherd99/abacus-BSE that referenced this pull request Mar 31, 2025
…ull size of Hamiltonian in multi-k case (deepmodeling#5785)

* renew deepks unit test

* renew deepks unit test

* reduce deepks-unitest's dependence on other modules

* fix the bug of "Couldn't find orbital files for descriptor"

* fix bugs of read orbitals and init orb&ParaO

* fix the bug of DMR ; change the names of files being compared

* fix: calculate H_V_delta

* fix : cal_H_V_delta for check_e_deltabands

* fix : Add pointer for GridD in deepks operator. use Test_Deepks::GridD in deepks unit test instead of GlobalC::GridD

* fix : modify deepks operator to support insert_pair in initilize_HR in multi-k

* fix : undefined reference of test_hsolver_sdft when using mpiicpx and icpx

* update intergrate test result. Note modification in deepks operator for complete atom_pairs will cause time increase in multi-k cases.

* remove some useless code

* remove nnr in deepks unittest

* fix: cutoff for overlap_orb_alpha_ in deepks test

* fix : psialpha ref of gamma only case , just change the order of neighour atoms

* fix uninitilized nmax_total in unitcell; simplize cal_descriptor

* restore deepks unit test in gamma-only case

* modify STRU of SiO2 in deepks test to reduce the number of neighbor atoms, thus reducing much time

* fix a bug when merge

* fix : setup nonlocal for ucell

* rename phialpha_ref.dat for multi-k

* modify STRU of H2O_multik in deepks test to reduce the number of neighbor atoms, thus reducing much time

* fix : add module paw dependence in deepks unittest

* fix : rename descriptor file

* fix : modify result.ref for SiO2. The privious modification don't change totaldes wrongly.

* update CMakeLists.txt
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Compile & CICD & Docs & Dependencies Issues related to compiling ABACUS Machine Learning Issues related to the DeePKS Refactor Refactor ABACUS codes

Projects

None yet

Development

Successfully merging this pull request may close these issues.

The unittests for DeePKS are not being compiled.

3 participants