Skip to content

Commit

Permalink
update
Browse files Browse the repository at this point in the history 
update

Fix default LOW_PREC
  • Loading branch information
@lidanqing-intel
lidanqing-intel committed May 18, 2022
1 parent 7ec293a commit 7c53dd9
Show file tree
Hide file tree
Showing 6 changed files with 74 additions and 73 deletions.
3 changes: 2 additions & 1 deletion source/CMakeLists.txt
View file
Original file line number Diff line number Diff line change
Expand Up @@ -171,13 +171,14 @@ if (DEFINED FLOAT_PREC)
string ( TOLOWER ${FLOAT_PREC} lower_float_prec )
if (lower_float_prec STREQUAL "high")
set(PREC_DEF "-DHIGH_PREC")
add_definitions(${PREC_DEF})
else ()
set(PREC_DEF "")
endif ()
else ()
set(PREC_DEF "-DHIGH_PREC")
add_definitions(${PREC_DEF})
endif()
add_definitions (${PREC_DEF})

# find openmp
find_package(OpenMP)
Expand Down
12 changes: 6 additions & 6 deletions source/api_cc/include/common.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -18,14 +18,14 @@ typedef std::string STRINGTYPE;
#endif

#ifdef HIGH_PREC
//typedef double VALUETYPE;
//typedef double ENERGYTYPE;
typedef float VALUETYPE;
typedef float ENERGYTYPE;
typedef double VALUETYPE;
typedef double ENERGYTYPE;
// typedef float VALUETYPE;
// typedef float ENERGYTYPE;
#else
typedef float VALUETYPE;
//typedef double ENERGYTYPE;
typedef float ENERGYTYPE;
typedef double ENERGYTYPE;
// typedef float ENERGYTYPE;
#endif

struct NeighborListData
Expand Down
32 changes: 16 additions & 16 deletions source/api_cc/src/PaddleDeepPot.cc
View file
Original file line number Diff line number Diff line change
Expand Up @@ -41,7 +41,8 @@ paddle_run_model(ENERGYTYPE & dener,
}
predictor->Run();
// Deal with output
std::vector<ENERGYTYPE> oe;
// std::vector<deepmd::ENERGYTYPE> oe;
std::vector<float> oe;
std::vector<VALUETYPE> of, oav;
auto output_names = predictor->GetOutputNames();

Expand All @@ -66,7 +67,7 @@ paddle_run_model(ENERGYTYPE & dener,
oav.resize(out_v_num);
out_v_t->CopyToCpu(oav.data());

dener = oe[0];
dener = static_cast<deepmd::ENERGYTYPE>(oe[0]);
std::vector<VALUETYPE> dforce (3 * nall, 0.0);
dvirial.resize(9);
for (int i = 0; i < nall * 3; ++i) {
Expand Down Expand Up @@ -123,7 +124,8 @@ paddle_run_model (deepmd::ENERGYTYPE & dener,
predictor->Run();
// {"o_energy", "o_force", "o_atom_energy", "o_atom_virial"},
// Deal with output
std::vector<deepmd::ENERGYTYPE> oe;
// std::vector<deepmd::ENERGYTYPE> oe;
std::vector<float> oe;
std::vector<deepmd::VALUETYPE> of, oav, oae;
auto output_names = predictor->GetOutputNames();

Expand Down Expand Up @@ -155,7 +157,7 @@ paddle_run_model (deepmd::ENERGYTYPE & dener,
oav.resize(out_av_num);
out_av_t->CopyToCpu(oav.data());

dener = oe[0];
dener = static_cast<deepmd::ENERGYTYPE>(oe[0]);
std::vector<VALUETYPE> dforce (3 * nall);
std::vector<VALUETYPE> datom_energy (nall, 0);
std::vector<VALUETYPE> datom_virial (9 * nall);
Expand Down Expand Up @@ -583,15 +585,15 @@ compute (std::vector<ENERGYTYPE> & all_energy,
assert (nloc == ret[ii]);
}
for (unsigned ii = 0; ii < numb_models; ++ii) {
// paddle_run_model (all_energy[ii], all_force[ii], all_virial[ii], all_atom_energy[ii], all_atom_virial[ii], sessions[ii], predictor_, atommap, nghost);
paddle_run_model (all_energy[ii], all_force[ii], all_virial[ii], all_atom_energy[ii], all_atom_virial[ii], predictors[ii], atommap, nghost);
paddle_run_model (all_energy[ii], all_force[ii], all_virial[ii], all_atom_energy[ii], all_atom_virial[ii], predictors[ii], atommap, nghost);
}
}

#ifndef HIGH_PREC
void
PaddleDeepPotModelDevi::
compute_avg (VALUETYPE & dener,
const std::vector<VALUETYPE > & all_energy)
compute_avg (ENERGYTYPE & dener,
const std::vector<ENERGYTYPE >& all_energy)
{
assert (all_energy.size() == numb_models);
if (numb_models == 0) return;
Expand All @@ -600,14 +602,14 @@ compute_avg (VALUETYPE & dener,
for (unsigned ii = 0; ii < numb_models; ++ii){
dener += all_energy[ii];
}
dener /= (VALUETYPE)(numb_models);
dener /= (ENERGYTYPE)(numb_models);
}
#endif

#ifndef HIGH_PREC
void
DeepPotModelDevi::
compute_avg (ENERGYTYPE & dener,
const std::vector<ENERGYTYPE >& all_energy)
PaddleDeepPotModelDevi::
compute_avg (VALUETYPE & dener,
const std::vector<VALUETYPE > & all_energy)
{
assert (all_energy.size() == numb_models);
if (numb_models == 0) return;
Expand All @@ -616,9 +618,8 @@ compute_avg (ENERGYTYPE & dener,
for (unsigned ii = 0; ii < numb_models; ++ii){
dener += all_energy[ii];
}
dener /= (ENERGYTYPE)(numb_models);
dener /= (VALUETYPE)(numb_models);
}
#endif

void
PaddleDeepPotModelDevi::
Expand All @@ -642,7 +643,6 @@ compute_avg (std::vector<VALUETYPE> & avg,
}
}


void
PaddleDeepPotModelDevi::
compute_std_e (std::vector<VALUETYPE> & std,
Expand Down
2 changes: 1 addition & 1 deletion source/lmp/env.sh.in
View file
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
DEEPMD_ROOT=@CMAKE_INSTALL_PREFIX@

NNP_INC=" -std=c++14 @PREC_DEF@ @TTM_DEF@ @OLD_LMP_PPPM_DEF@ -I$DEEPMD_ROOT/include -I$PADDLE_ROOT"
NNP_INC=" -std=c++14 @PREC_DEF@ @TTM_DEF@ @OLD_LMP_PPPM_DEF@ -I$DEEPMD_ROOT/include -I$paddle_root"
NNP_PATH=" -L$DEEPMD_ROOT/lib"
NNP_LIB=" -Wl,--no-as-needed -l@LIB_DEEPMD_CC@ -l@LIB_DEEPMD@ -Wl,-rpath=$DEEPMD_ROOT/lib -lpd_infer_custom_op"
94 changes: 47 additions & 47 deletions source/lmp/pair_deepmd.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -131,10 +131,10 @@ std::vector<std::string> PairDeepMD::get_params_file_content(const std::vector<s
}

static void
ana_st (float & max,
float & min,
float & sum,
const vector<float> & vec,
ana_st (double & max,
double & min,
double & sum,
const vector<double> & vec,
const int & nloc)
{
if (nloc == 0) return;
Expand All @@ -151,8 +151,8 @@ ana_st (float & max,
static void
make_uniform_aparam(
#ifdef HIGH_PREC
vector<float > & daparam,
const vector<float > & aparam,
vector<double > & daparam,
const vector<double > & aparam,
const int & nlocal
#else
vector<float > & daparam,
Expand All @@ -173,7 +173,7 @@ make_uniform_aparam(
#ifdef USE_TTM
void PairDeepMD::make_ttm_aparam(
#ifdef HIGH_PREC
vector<float > & daparam
vector<double > & daparam
#else
vector<float > & daparam
#endif
Expand All @@ -196,18 +196,18 @@ void PairDeepMD::make_ttm_aparam(
int nxnodes = nnodes[0];
int nynodes = nnodes[1];
int nznodes = nnodes[2];
float *** const T_electron = ttm_fix->get_T_electron();
float dx = domain->xprd/nxnodes;
float dy = domain->yprd/nynodes;
float dz = domain->zprd/nynodes;
double *** const T_electron = ttm_fix->get_T_electron();
double dx = domain->xprd/nxnodes;
double dy = domain->yprd/nynodes;
double dz = domain->zprd/nynodes;
// resize daparam
daparam.resize(nlocal);
// loop over atoms to assign aparam
for (int ii = 0; ii < nlocal; ii++) {
if (mask[ii] & ttm_fix->groupbit) {
float xscale = static_cast<float>((x[ii][0] - domain->boxlo[0])/domain->xprd);
float yscale = static_cast<float>((x[ii][1] - domain->boxlo[1])/domain->yprd);
float zscale = static_cast<float>((x[ii][2] - domain->boxlo[2])/domain->zprd);
double xscale = static_cast<double>((x[ii][0] - domain->boxlo[0])/domain->xprd);
double yscale = static_cast<double>((x[ii][1] - domain->boxlo[1])/domain->yprd);
double zscale = static_cast<double>((x[ii][2] - domain->boxlo[2])/domain->zprd);
int ixnode = static_cast<int>(xscale*nxnodes);
int iynode = static_cast<int>(yscale*nynodes);
int iznode = static_cast<int>(zscale*nznodes);
Expand Down Expand Up @@ -300,13 +300,13 @@ void PairDeepMD::compute(int eflag, int vflag)
dtype[ii] = type[ii] - 1;
}

float dener (0);
vector<float > dforce (nall * 3);
vector<float > dvirial (9, 0);
vector<float > dcoord (nall * 3, 0.);
vector<float > dbox (9, 0) ;
double dener (0);
vector<double > dforce (nall * 3);
vector<double > dvirial (9, 0);
vector<double > dcoord (nall * 3, 0.);
vector<double > dbox (9, 0) ;
#ifdef HIGH_PREC
vector<float > daparam;
vector<double > daparam;
#else
vector<float > daparam;
#endif
Expand Down Expand Up @@ -363,7 +363,7 @@ void PairDeepMD::compute(int eflag, int vflag)
for (unsigned dd = 0; dd < dbox.size(); ++dd) dbox_[dd] = dbox[dd];
vector<float> dforce_(dforce.size(), 0);
vector<float> dvirial_(dvirial.size(), 0);
float dener_ = 0;
double dener_ = 0;
deep_pot.compute (dener_, dforce_, dvirial_, dcoord_, dtype, dbox_, nghost, lmp_list, ago, fparam, daparam);
for (unsigned dd = 0; dd < dforce.size(); ++dd) dforce[dd] = dforce_[dd];
for (unsigned dd = 0; dd < dvirial.size(); ++dd) dvirial[dd] = dvirial_[dd];
Expand All @@ -372,8 +372,8 @@ void PairDeepMD::compute(int eflag, int vflag)
}
// do atomic energy and virial
else {
vector<float > deatom (nall * 1, 0);
vector<float > dvatom (nall * 9, 0);
vector<double > deatom (nall * 1, 0);
vector<double > dvatom (nall * 9, 0);
#ifdef HIGH_PREC
deep_pot.compute (dener, dforce, dvirial, deatom, dvatom, dcoord, dtype, dbox, nghost, lmp_list, ago, fparam, daparam);
#else
Expand All @@ -385,7 +385,7 @@ void PairDeepMD::compute(int eflag, int vflag)
vector<float> dvirial_(dvirial.size(), 0);
vector<float> deatom_(dforce.size(), 0);
vector<float> dvatom_(dforce.size(), 0);
float dener_ = 0;
double dener_ = 0;
deep_pot.compute (dener_, dforce_, dvirial_, deatom_, dvatom_, dcoord_, dtype, dbox_, nghost, lmp_list, ago, fparam, daparam);
for (unsigned dd = 0; dd < dforce.size(); ++dd) dforce[dd] = dforce_[dd];
for (unsigned dd = 0; dd < dvirial.size(); ++dd) dvirial[dd] = dvirial_[dd];
Expand All @@ -409,12 +409,12 @@ void PairDeepMD::compute(int eflag, int vflag)
}
}
else if (multi_models_mod_devi) {
vector<float > deatom (nall * 1, 0);
vector<float > dvatom (nall * 9, 0);
vector<double > deatom (nall * 1, 0);
vector<double > dvatom (nall * 9, 0);
#ifdef HIGH_PREC
vector<float> all_energy;
vector<vector<float>> all_virial;
vector<vector<float>> all_atom_energy;
vector<double> all_energy;
vector<vector<double>> all_virial;
vector<vector<double>> all_atom_energy;
vector<vector<float>> all_atom_virial;
deep_pot_model_devi.compute(all_energy, all_force, all_virial, all_atom_energy, all_atom_virial, dcoord, dtype, dbox, nghost, lmp_list, ago, fparam, daparam);
// deep_pot_model_devi.compute_avg (dener, all_energy);
Expand All @@ -436,8 +436,8 @@ void PairDeepMD::compute(int eflag, int vflag)
vector<float> dvirial_(dvirial.size(), 0);
vector<float> deatom_(dforce.size(), 0);
vector<float> dvatom_(dforce.size(), 0);
float dener_ = 0;
vector<float> all_energy_;
double dener_ = 0;
vector<double> all_energy_;
vector<vector<float>> all_force_;
vector<vector<float>> all_virial_;
vector<vector<float>> all_atom_energy_;
Expand Down Expand Up @@ -485,9 +485,9 @@ void PairDeepMD::compute(int eflag, int vflag)
if (newton_pair) {
comm->reverse_comm_pair(this);
}
vector<float> std_f;
vector<double> std_f;
#ifdef HIGH_PREC
vector<float> tmp_avg_f;
vector<double> tmp_avg_f;
deep_pot_model_devi.compute_avg (tmp_avg_f, all_force);
deep_pot_model_devi.compute_std_f (std_f, tmp_avg_f, all_force);
if (out_rel == 1){
Expand All @@ -508,19 +508,19 @@ void PairDeepMD::compute(int eflag, int vflag)
deep_pot_model_devi.compute_relative_std_f (std_f_, tmp_avg_f_, eps);
}
#endif
float min = numeric_limits<float>::max(), max = 0, avg = 0;
double min = numeric_limits<double>::max(), max = 0, avg = 0;
ana_st(max, min, avg, std_f, nlocal);
int all_nlocal = 0;
MPI_Reduce (&nlocal, &all_nlocal, 1, MPI_INT, MPI_SUM, 0, world);
float all_f_min = 0, all_f_max = 0, all_f_avg = 0;
double all_f_min = 0, all_f_max = 0, all_f_avg = 0;
MPI_Reduce (&min, &all_f_min, 1, MPI_DOUBLE, MPI_MIN, 0, world);
MPI_Reduce (&max, &all_f_max, 1, MPI_DOUBLE, MPI_MAX, 0, world);
MPI_Reduce (&avg, &all_f_avg, 1, MPI_DOUBLE, MPI_SUM, 0, world);
all_f_avg /= float(all_nlocal);
all_f_avg /= double(all_nlocal);
// std energy
vector<float > std_e;
vector<double > std_e;
#ifdef HIGH_PREC
vector<float > tmp_avg_e;
vector<double > tmp_avg_e;
deep_pot_model_devi.compute_avg (tmp_avg_e, all_atom_energy);
deep_pot_model_devi.compute_std_e (std_e, tmp_avg_e, all_atom_energy);
#else
Expand All @@ -531,20 +531,20 @@ void PairDeepMD::compute(int eflag, int vflag)
for (int dd = 0; dd < std_e_.size(); ++dd) std_e[dd] = std_e_[dd];
#endif
max = avg = 0;
min = numeric_limits<float>::max();
min = numeric_limits<double>::max();
ana_st(max, min, avg, std_e, nlocal);
float all_e_min = 0, all_e_max = 0, all_e_avg = 0;
double all_e_min = 0, all_e_max = 0, all_e_avg = 0;
MPI_Reduce (&min, &all_e_min, 1, MPI_DOUBLE, MPI_MIN, 0, world);
MPI_Reduce (&max, &all_e_max, 1, MPI_DOUBLE, MPI_MAX, 0, world);
MPI_Reduce (&avg, &all_e_avg, 1, MPI_DOUBLE, MPI_SUM, 0, world);
all_e_avg /= float(all_nlocal);
all_e_avg /= double(all_nlocal);
// // total e
// vector<float > sum_e(numb_models, 0.);
// vector<double > sum_e(numb_models, 0.);
// MPI_Reduce (&all_energy[0], &sum_e[0], numb_models, MPI_DOUBLE, MPI_SUM, 0, world);
if (rank == 0) {
// float avg_e = 0;
// double avg_e = 0;
// deep_pot_model_devi.compute_avg(avg_e, sum_e);
// float std_e_1 = 0;
// double std_e_1 = 0;
// deep_pot_model_devi.compute_std(std_e_1, avg_e, sum_e);
fp << setw(12) << update->ntimestep
<< " " << setw(18) << all_e_max
Expand Down Expand Up @@ -582,7 +582,7 @@ void PairDeepMD::compute(int eflag, int vflag)
for (unsigned dd = 0; dd < dbox.size(); ++dd) dbox_[dd] = dbox[dd];
vector<float> dforce_(dforce.size(), 0);
vector<float> dvirial_(dvirial.size(), 0);
float dener_ = 0;
double dener_ = 0;
deep_pot.compute (dener_, dforce_, dvirial_, dcoord_, dtype, dbox_);
for (unsigned dd = 0; dd < dforce.size(); ++dd) dforce[dd] = dforce_[dd];
for (unsigned dd = 0; dd < dvirial.size(); ++dd) dvirial[dd] = dvirial_[dd];
Expand Down Expand Up @@ -890,7 +890,7 @@ double PairDeepMD::init_one(int i, int j)

/* ---------------------------------------------------------------------- */

int PairDeepMD::pack_reverse_comm(int n, int first, float *buf)
int PairDeepMD::pack_reverse_comm(int n, int first, double *buf)
{
int i,m,last;

Expand All @@ -908,7 +908,7 @@ int PairDeepMD::pack_reverse_comm(int n, int first, float *buf)

/* ---------------------------------------------------------------------- */

void PairDeepMD::unpack_reverse_comm(int n, int *list, float *buf)
void PairDeepMD::unpack_reverse_comm(int n, int *list, double *buf)
{
int i,j,m;

Expand Down
4 changes: 2 additions & 2 deletions source/lmp/pair_deepmd.h.in
View file
Original file line number Diff line number Diff line change
Expand Up @@ -43,8 +43,8 @@ class PairDeepMD : public Pair {
virtual void coeff(int, char **);
void init_style();
double init_one(int i, int j);
int pack_reverse_comm(int, int, float *);
void unpack_reverse_comm(int, int *, float *);
int pack_reverse_comm(int, int, double *);
void unpack_reverse_comm(int, int *, double *);
void print_summary(const std::string pre) const;
int get_node_rank();
std::string get_prog_file_content(const std::string & model);
Expand Down

0 comments on commit 7c53dd9

Please sign in to comment.