Merge branch 'deepmodeling:develop' into develop

Author: 1041176461

.github/workflows/build_test_cmake.yml

@@ -17,10 +17,10 @@ jobs:
             name: "Build with Intel toolchain"
 
           - tag: gnu
-            build_args: "-DENABLE_LIBXC=1 -DENABLE_DEEPKS=1 -DENABLE_LIBRI=1 -DENABLE_PAW=1"
+            build_args: "-DENABLE_LIBXC=1 -DENABLE_DEEPKS=1 -DENABLE_MLKEDF=1 -DENABLE_LIBRI=1 -DENABLE_PAW=1"
             name: "Build extra components with GNU toolchain"
           - tag: intel
-            build_args: "-DENABLE_LIBXC=1 -DENABLE_DEEPKS=1 -DENABLE_LIBRI=1"
+            build_args: "-DENABLE_LIBXC=1 -DENABLE_DEEPKS=1 -DENABLE_MLKEDF=1 -DENABLE_LIBRI=1"
             name: "Build extra components with Intel toolchain"
 
           - tag: cuda

.github/workflows/test.yml

@@ -31,7 +31,7 @@ jobs:
 
       - name: Configure
         run: |
-          cmake -B build -DBUILD_TESTING=ON -DENABLE_DEEPKS=ON -DENABLE_LIBXC=ON -DENABLE_LIBRI=ON -DENABLE_PAW=ON -DENABLE_GOOGLEBENCH=ON -DENABLE_RAPIDJSON=ON  -DCMAKE_EXPORT_COMPILE_COMMANDS=1
+          cmake -B build -DBUILD_TESTING=ON -DENABLE_DEEPKS=ON -DENABLE_MLKEDF=ON -DENABLE_LIBXC=ON -DENABLE_LIBRI=ON -DENABLE_PAW=ON -DENABLE_GOOGLEBENCH=ON -DENABLE_RAPIDJSON=ON  -DCMAKE_EXPORT_COMPILE_COMMANDS=1
 
       - uses: pre-commit/[email protected]
         with:

CMakeLists.txt

@@ -12,6 +12,7 @@ project(
 
 option(ENABLE_LCAO "Enable LCAO calculation." ON)
 option(ENABLE_DEEPKS "Enable DeePKS functionality" OFF)
+option(ENABLE_MLKEDF "Enable Machine Learning based KEDF for OFDFT" OFF)
 option(ENABLE_LIBXC "Enable LibXC functionality" OFF)
 option(USE_CUDA "Enable support to CUDA for ABACUS." OFF)
 option(ENABLE_FLOAT_FFTW "Enable support to single precision FFTW library." OFF)
@@ -40,6 +41,7 @@ option(ENABLE_CNPY "Enable cnpy usage." OFF)
 option(ENABLE_PEXSI "Enable support for PEXSI." OFF)
 option(ENABLE_CUSOLVERMP "Enable cusolvermp." OFF)
 option(USE_DSP "Enable DSP usage." OFF)
+option(USE_CUDA_ON_DCU "Enable CUDA on DCU" OFF)
 
 # enable json support
 if(ENABLE_RAPIDJSON)
@@ -126,6 +128,10 @@ if (USE_DSP)
   set(ABACUS_BIN_NAME abacus_dsp)
 endif()
 
+if (USE_CUDA_ON_DCU)
+  add_compile_definitions(__CUDA_ON_DCU)
+endif()
+
 list(APPEND CMAKE_MODULE_PATH ${CMAKE_CURRENT_SOURCE_DIR}/cmake)
 
 if(ENABLE_COVERAGE)
@@ -421,10 +427,8 @@ else()
   include_directories(${FFTW3_INCLUDE_DIRS})
   list(APPEND math_libs FFTW3::FFTW3 LAPACK::LAPACK BLAS::BLAS)
 
-  if(ENABLE_LCAO)
-    find_package(ScaLAPACK REQUIRED)
-    list(APPEND math_libs ScaLAPACK::ScaLAPACK)
-  endif()
+  find_package(ScaLAPACK REQUIRED)
+  list(APPEND math_libs ScaLAPACK::ScaLAPACK)
 
   if(USE_OPENMP)
     list(APPEND math_libs FFTW3::FFTW3_OMP)
@@ -466,10 +470,29 @@ if(ENABLE_DEEPKS)
   add_compile_definitions(__DEEPKS)
 endif()
 
+if(ENABLE_MLKEDF)
+  target_link_libraries(${ABACUS_BIN_NAME} hamilt_mlkedf)
+
+  find_path(libnpy_SOURCE_DIR npy.hpp HINTS ${libnpy_INCLUDE_DIR})
+  if(NOT libnpy_SOURCE_DIR)
+    include(FetchContent)
+    FetchContent_Declare(
+      libnpy
+      GIT_REPOSITORY https://github.com/llohse/libnpy.git
+      GIT_SHALLOW TRUE
+      GIT_PROGRESS TRUE)
+    FetchContent_MakeAvailable(libnpy)
+  else()
+    include_directories(${libnpy_INCLUDE_DIR})
+  endif()
+  include_directories(${libnpy_SOURCE_DIR}/include)
+  add_compile_definitions(__MLKEDF)
+endif()
+
 # Torch uses outdated components to detect CUDA arch, causing failure on
 # latest CUDA kits. Set CMake variable TORCH_CUDA_ARCH_LIST in the form of
 # "major.minor" if required.
-if(ENABLE_DEEPKS OR DEFINED Torch_DIR)
+if(ENABLE_DEEPKS OR ENABLE_MLKEDF OR DEFINED Torch_DIR)
   find_package(Torch REQUIRED)
   if(NOT Torch_VERSION VERSION_LESS "2.1.0")
     set_if_higher(CMAKE_CXX_STANDARD 17)
@@ -713,19 +736,20 @@ target_link_libraries(
   esolver
   vdw
   device
-  container)
+  container
+  dftu
+  deltaspin)
 if(ENABLE_LCAO)
   target_link_libraries(
     ${ABACUS_BIN_NAME}
     hamilt_lcao
     tddft
     orb
     gint
-    dftu
     hcontainer
-    deltaspin
     numerical_atomic_orbitals
-    lr)
+    lr
+    rdmft)
   if(USE_ELPA)
     target_link_libraries(${ABACUS_BIN_NAME} genelpa)
   endif()

docs/CITATIONS.md

@@ -33,3 +33,9 @@ The following references are required to be cited when using ABACUS. Specificall
     Peize Lin, Xinguo Ren, and Lixin He. "Efficient Hybrid Density Functional Calculations for Large Periodic Systems Using Numerical Atomic Orbitals." Journal of Chemical Theory and Computation 2021, 17(1), 222–239.
 
     Peize Lin, Xinguo Ren, and Lixin He. "Accuracy of Localized Resolution of the Identity in Periodic Hybrid Functional Calculations with Numerical Atomic Orbitals." Journal of Physical Chemistry Letters 2020, 11, 3082-3088.
+
+- **If ML-KEDF is used:**
+
+    Sun, Liang, and Mohan Chen. "Machine learning based nonlocal kinetic energy density functional for simple metals and alloys." Physical Review B 109.11 (2024): 115135.
+
+    Sun, Liang, and Mohan Chen. "Multi-channel machine learning based nonlocal kinetic energy density functional for semiconductors." Electronic Structure 6.4 (2024): 045006.

docs/advanced/elec_properties/hs_matrix.md

@@ -61,6 +61,11 @@ We also offer the option of only calculating the overlap matrix without running
 
 A file named `SR.csr` will be generated in the working directory, which contains the overlap matrix.
 
+> When `nspin` is set to 1 or 2, the dimension of the overlap matrix is nlocal $\times$ nlocal, where nlocal is the total number of numerical atomic orbitals. 
+These numerical atomic orbitals are ordered from outer to inner loop as atom, angular quantum number $l$, zeta (multiple radial orbitals corresponding to each $l$), and magnetic quantum number $m$. 
+When `nspin` is set to 4, the dimension of the overlap matrix is (2 $\times$ nlocal) $\times$ (2 $\times$ nlocal). In this case, the numerical atomic orbitals are ordered from outer to inner loop as atom, angular quantum number $l$, zeta (multiple radial orbitals corresponding to each $l$), magnetic quantum number $m$, and npol (index of spin, ranges from 0 to 1).
+
+
 ## examples
 We provide [examples](https://github.com/deepmodeling/abacus-develop/tree/develop/examples/matrix_hs) of outputting the matrices. There are four examples:
 

docs/advanced/elec_properties/position_matrix.md

@@ -18,4 +18,7 @@ Each file or each section of the appended file starts with "STEP: " followed by
 
 Each block here contains the matrix for the corresponding cell. There are three columns in each block, giving the matrix elements in x, y, z directions, respectively. There are altogether nbasis * nbasis lines in each block, which emulates the matrix elements.
 
-In molecular dynamics (MD) calculations, if [out_app_flag](../input_files/input-main.md#out_app_flag) is set to true, then `data-rR-tr` is written in an append manner. Otherwise, output files will be put in a separate directory, `matrix`, and named as `$x`_data-rR-tr, where `$x` is the number of MD step. In addition, The output frequency is controlled by [out_interval](../input_files/input-main.md#out_interval). For example, if we are running a 10-step MD with out_interval = 3, then `$x` will be 0, 3, 6, and 9.
+In molecular dynamics (MD) calculations, if [out_app_flag](../input_files/input-main.md#out_app_flag) is set to true, then `data-rR-tr` is written in an append manner. Otherwise, output files will be put in a separate directory, `matrix`, and named as `$x`_data-rR-tr, where `$x` is the number of MD step. In addition, the output frequency is controlled by [out_interval](../input_files/input-main.md#out_interval). For example, if we are running a 10-step MD with out_interval = 3, then `$x` will be 0, 3, 6, and 9.
+
+## get_S
+We also offer the option of only calculating the position matrix without running SCF. For that purpose, in `INPUT` file we need to set the keyword [calculation](../input_files/input-main.md#calculation) to `get_S`, and [out_mat_r](../input_files/input-main.md#out_mat_r) to `true`.