Merge branch 'develop' into modularize_1

Author: Critsium-xy

docs/advanced/input_files/input-main.md

@@ -65,6 +65,8 @@
     - [nupdown](#nupdown)
     - [dft\_functional](#dft_functional)
     - [xc\_temperature](#xc_temperature)
+    - [xc\_exch\_ext](#xc_exch_ext)
+    - [xc\_corr\_ext](#xc_corr_ext)
     - [pseudo\_rcut](#pseudo_rcut)
     - [pseudo\_mesh](#pseudo_mesh)
     - [nspin](#nspin)
@@ -1060,6 +1062,26 @@ calculations.
 - **Default**: 0.0
 - **Unit**: Ry
 
+### xc_exch_ext
+
+- **Type**: Integer Real ...
+- **Description**: Customized parameterization on the exchange part of XC functional. The first value should be the LibXC ID of the original functional, and latter values are external parameters. Default values are those of Perdew-Burke-Ernzerhof (PBE) functional. For more information on LibXC ID of functionals, please refer to [LibXC](https://libxc.gitlab.io/functionals/). For parameters of functionals of interest, please refer to the source code of LibXC, such as PBE functional interface in LibXC: [gga_x_pbe.c](https://gitlab.com/libxc/libxc/-/blob/7.0.0/src/gga_x_pbe.c).
+- **Default**: 101 0.8040 0.2195149727645171
+- **Note**:
+  1. Solely setting this keyword will take no effect on XC functionals. One should also set `dft_functional` to corresponding functional to apply the customized parameterization. For example, if you want to use the PBE functional with customized parameters, you should set `dft_functional` to `GGA_X_PBE+GGA_C_PBE` and `xc_exch_ext` to `101 0.8040 0.2195149727645171`.
+  2. For functionals that do not have separate exchange and correlation parts, such as HSE06 whose corresponding LibXC notation is `HYB_GGA_XC_HSE06` and LibXC id is 428, you can set either `xc_exch_ext` or `xc_corr_ext` to `428 0.25 0.11 0.11` (which means 25% Hartree-Fock fraction, 0.11 as range-seperation) and leave the other one unset.
+  3. Presently this feature can only support parameterization on **one** exchange functional.
+
+### xc_corr_ext
+
+- **Type**: Integer Real ...
+- **Description**: Customized parameterization on the correlation part of XC functional. The first value should be the LibXC ID of the original functional, and latter values are external parameters. Default values are those of Perdew-Burke-Ernzerhof (PBE) functional. For more information on LibXC ID of functionals, please refer to [LibXC](https://libxc.gitlab.io/functionals/). For parameters of functionals of interest, please refer to the source code of LibXC, such as PBE functional interface in LibXC: [gga_c_pbe.c](https://gitlab.com/libxc/libxc/-/blob/7.0.0/src/gga_c_pbe.c).
+- **Default**: 130 0.06672455060314922 0.031090690869654895034 1.0
+- **Note**:
+  1. Solely setting this keyword will take no effect on XC functionals. One should also set `dft_functional` to corresponding functional to apply the customized parameterization. For example, if you want to use the PBE functional with customized parameters, you should set `dft_functional` to `GGA_X_PBE+GGA_C_PBE` and `xc_corr_ext` to `130 0.06672455060314922 0.031090690869654895034 1.0`.
+  2. For functionals that do not have separate exchange and correlation parts, such as HSE06 whose corresponding LibXC notation is `HYB_GGA_XC_HSE06` and LibXC id is 428, you can set either `xc_exch_ext` or `xc_corr_ext` to `428 0.25 0.11 0.11` (which means 25% Hartree-Fock fraction, 0.11 as range-seperation) and leave the other one unset.
+  3. Presently this feature can only support parameterization on **one** correlation functional.
+
 ### pseudo_rcut
 
 - **Type**: Real
@@ -2378,8 +2400,29 @@ Warning: this function is not robust enough for the current version. Please try
 ### of_ml_gene_data
 
 - **Type**: Boolean
-- **Availability**: OFDFT
-- **Description**: Generate training data or not.
+- **Availability**: Used only for KSDFT with plane wave basis
+- **Description**: Controls the generation of machine learning training data. When enabled, training data in `.npy` format will be saved in the directory `OUT.${suffix}/MLKEDF_Descriptors/`. The generated descriptors are categorized as follows:
+  - Local/Semilocal Descriptors. Files are named as `{var}.npy`, where `{var}` corresponds to the descriptor type:
+    - `gamma`: Enabled by [of_ml_gamma](#of_ml_gamma)  
+    - `p`: Enabled by [of_ml_p](#of_ml_p)  
+    - `q`: Enabled by [of_ml_q](#of_ml_q)  
+    - `tanhp`: Enabled by [of_ml_tanhp](#of_ml_tanhp)  
+    - `tanhq`: Enabled by [of_ml_tanhq](#of_ml_tanhq)  
+  - Nonlocal Descriptors generated using kernels configured via [of_ml_nkernel](#of_ml_nkernel), [of_ml_kernel](#of_ml_kernel), and [of_ml_kernel_scaling](#of_ml_kernel_scaling). Files follow the naming convention `{var}_{kernel_type}_{kernel_scaling}.npy`, where `{kernel_type}` and `{kernel_scaling}` are specified by [of_ml_kernel](#of_ml_kernel), and [of_ml_kernel_scaling](#of_ml_kernel_scaling), respectively, and `{val}` denotes the kind of the descriptor, including 
+      - `gammanl`: Enabled by [of_ml_gammanl](#of_ml_gammanl)  
+      - `pnl`: Enabled by [of_ml_pnl](#of_ml_pnl)  
+      - `qnl`: Enabled by [of_ml_qnl](#of_ml_qnl)  
+      - `xi`: Enabled by [of_ml_xi](#of_ml_xi)  
+      - `tanhxi`: Enabled by [of_ml_tanhxi](#of_ml_tanhxi)  
+      - `tanhxi_nl`: Enabled by [of_ml_tanhxi_nl](#of_ml_tanhxi_nl)  
+      - `tanh_pnl`: Enabled by [of_ml_tanh_pnl](#of_ml_tanh_pnl)  
+      - `tanh_qnl`: Enabled by [of_ml_tanh_qnl](#of_ml_tanh_qnl)  
+      - `tanhp_nl`: Enabled by [of_ml_tanhp_nl](#of_ml_tanhp_nl)  
+      - `tanhq_nl`: Enabled by [of_ml_tanhq_nl](#of_ml_tanhq_nl) 
+  - Training Targets, including key quantum mechanical quantities:
+    - `enhancement.npy`: Pauli energy enhancement factor $T_\theta/T_{\rm{TF}}$, where $T_{\rm{TF}}$ is the Thomas-Fermi functional  
+    - `pauli.npy`: Pauli potential $V_\theta$
+    - `veff.npy`: Effective potential
 - **Default**: False
 
 ### of_ml_device

source/Makefile.Objects

@@ -202,6 +202,7 @@ OBJS_CELL=atom_pseudo.o\
 
 OBJS_DEEPKS=LCAO_deepks.o\
         deepks_basic.o\
+        deepks_check.o\
         deepks_descriptor.o\
         deepks_force.o\
         deepks_fpre.o\

source/module_base/global_file.cpp

@@ -153,6 +153,47 @@ void ModuleBase::Global_File::make_dir_out(
 #endif
     }
 
+    if(PARAM.inp.of_ml_gene_data == 1)
+    {
+        int make_dir_descrip = 0;
+        std::string command1 =  "test -d " + PARAM.globalv.global_mlkedf_descriptor_dir + " || mkdir " + PARAM.globalv.global_mlkedf_descriptor_dir;
+
+        times = 0;
+        while(times<GlobalV::NPROC)
+        {
+            if(rank==times)
+            {
+                if ( system( command1.c_str() ) == 0 )
+                {
+                    std::cout << " MAKE THE MLKEDF DESCRIPTOR DIR    : " << PARAM.globalv.global_mlkedf_descriptor_dir << std::endl;
+                    make_dir_descrip = 1;
+                }
+                else
+                {
+                    std::cout << " PROC " << rank << " CAN NOT MAKE THE MLKEDF DESCRIPTOR DIR !!! " << std::endl;
+                    make_dir_descrip = 0;
+                }
+            }
+#ifdef __MPI
+            Parallel_Reduce::reduce_all(make_dir_descrip);
+#endif
+            if(make_dir_descrip > 0)
+            { 
+                break;
+            }
+            ++times;
+        }
+
+#ifdef __MPI
+        if(make_dir_descrip == 0)
+        {
+            std::cout << " CAN NOT MAKE THE MLKEDF DESCRIPTOR DIR......." << std::endl;
+            ModuleBase::QUIT();
+        }
+        MPI_Barrier(MPI_COMM_WORLD);
+#endif
+    }
+
     // mohan add 2010-09-12
     if(out_alllog)
     {

source/module_esolver/esolver_ks_pw.cpp

@@ -37,7 +37,7 @@
 #include <iostream>
 
 #ifdef __MLALGO
-#include "module_hamilt_pw/hamilt_ofdft/ml_data.h"
+#include "module_io/write_mlkedf_descriptors.h"
 #endif
 
 #include <ATen/kernels/blas.h>
@@ -1003,30 +1003,31 @@ void ESolver_KS_PW<T, Device>::after_all_runners(UnitCell& ucell)
     {
         this->pelec->pot->update_from_charge(&this->chr, &ucell);
 
-        ML_data ml_data;
-        ml_data.set_para(this->chr.nrxx,
-                         PARAM.inp.nelec,
-                         PARAM.inp.of_tf_weight,
-                         PARAM.inp.of_vw_weight,
-                         PARAM.inp.of_ml_chi_p,
-                         PARAM.inp.of_ml_chi_q,
-                         PARAM.inp.of_ml_chi_xi,
-                         PARAM.inp.of_ml_chi_pnl,
-                         PARAM.inp.of_ml_chi_qnl,
-                         PARAM.inp.of_ml_nkernel,
-                         PARAM.inp.of_ml_kernel,
-                         PARAM.inp.of_ml_kernel_scaling,
-                         PARAM.inp.of_ml_yukawa_alpha,
-                         PARAM.inp.of_ml_kernel_file,
-                         ucell.omega,
-                         this->pw_rho);
-
-        ml_data.generateTrainData_KS(this->kspw_psi,
-                                     this->pelec,
-                                     this->pw_wfc,
-                                     this->pw_rho,
-                                     ucell,
-                                     this->pelec->pot->get_effective_v(0));
+        ModuleIO::Write_MLKEDF_Descriptors write_mlkedf_desc;
+        write_mlkedf_desc.cal_tool->set_para(this->chr.nrxx,
+                                            PARAM.inp.nelec,
+                                            PARAM.inp.of_tf_weight,
+                                            PARAM.inp.of_vw_weight,
+                                            PARAM.inp.of_ml_chi_p,
+                                            PARAM.inp.of_ml_chi_q,
+                                            PARAM.inp.of_ml_chi_xi,
+                                            PARAM.inp.of_ml_chi_pnl,
+                                            PARAM.inp.of_ml_chi_qnl,
+                                            PARAM.inp.of_ml_nkernel,
+                                            PARAM.inp.of_ml_kernel,
+                                            PARAM.inp.of_ml_kernel_scaling,
+                                            PARAM.inp.of_ml_yukawa_alpha,
+                                            PARAM.inp.of_ml_kernel_file,
+                                            ucell.omega,
+                                            this->pw_rho);
+
+        write_mlkedf_desc.generateTrainData_KS(PARAM.globalv.global_mlkedf_descriptor_dir,
+                                                this->kspw_psi,
+                                                this->pelec,
+                                                this->pw_wfc,
+                                                this->pw_rho,
+                                                ucell,
+                                                this->pelec->pot->get_effective_v(0));
     }
 #endif
 }