Skip to content

v3.0.0: Towards AI+DFT

Compare
Choose a tag to compare
@caic99 caic99 released this 30 Sep 09:39
· 2226 commits to develop since this release
7549d88

Release Note

This update focuses on new AI-assisted algorithms: the machine learning assisted functional model DeePKS, the feature of generating machine learning atomic interaction potential functions with DP-GEN, and the interface with DeepH which generates Hamiltonian based on machine learning methods. In addition, ABACUS implements new features such as Meta-GGA and Stochastic Density Functional Theory (SDFT). ABACUS has also made significant efficiency optimizations. In addition, a brand-new documentation website will further enhance the user experience.

Feature

  • feature: add the output of .cube file in LCAO out_chg calculation by @pxlxingliang in #1310
  • Feature: add xc_temperature input keyword for finite T functionals by @wenfei-li in #1227
  • feature: output H/S matrix into a sub dir in md calculation by @YuLiu98 in #1258

Test

Bug fix

Documentation

Our developers have elaborately prepared the all-new documentation website. Please refer to the "Full Changelog" part for changes on documentation.

Refactor

  • refactor : absorb 'INPUTs ' into INPUT when generating DeePKS projectors by @wenfei-li in #1244
  • refactor: add allocateRho func in elecstate module by @hongriTianqi in #1266
  • refactor: add the output of lacked INPUT parameters in OUT.XXX/INPUT by @pxlxingliang in #1232
  • Refactor: delete the useless INPUT parameters restart_mode and test_just_neighbor by @pxlxingliang in #1233
  • Refactor: Remove the global variable global_pseudo_type. by @sunliang98 in #1247
  • refactor:in the diago of dav, modify to diagonalize the subspace only on rank0 by @pxlxingliang in #1221

Full Changelog:

v2.3.5...v3.0.0