def main():
# main settings
# todo: has to be inherited from the sh file:
sim = 'sim' # simulation folder name
database_folder = 'db'
# todo: this must be inherited
bh5670 = 'bh5670' # the outermost folder in the scratch where all other data are put
# todo yaml or predict
my_offset = 15 # from the molecule to a vacuum. 15-20 is recommended!
cutoff = 500
rel_cutoff = 50
basis_set_file_name = 'BASIS_CC_AUG_RI_NEW' # RI5, 2-5 cc, all aug-cc, RIFIT-all
my_basis_sets = [
'aug-cc-pVDZ', 'aug-cc-pVTZ', 'aug-cc-pVQZ', 'aug-cc-pV5Z'
]
my_ri_basis_sets = [
'aug-cc-pVDZ-RIFIT', 'aug-cc-pVTZ-RIFIT', 'aug-cc-pVQZ-RIFIT',
'aug-cc-pV5Z-RIFIT'
]
debug = True
# end: main settings
dummy_run = False # does not invoke cp2k if true
# parser begin
parser = argparse.ArgumentParser(description='rank and num of cpus')
parser.add_argument('-rank') # array job number
parser.add_argument(
'-num_cpus') # number of cpus you request for every array job
args = parser.parse_args()
# parser end
# parsing input
threads = int(args.num_cpus) - 1 # cpus used to compute
rank = '{:0>6}'.format(
args.rank) # transform rank from '1' to '000001' format
prefix_xyz_file_name = 'dsgdb9nsd'
xyz_file_name = f'{prefix_xyz_file_name}_{rank}.xyz'
xyz_file_location = f'{prefix_xyz_file_name}/{xyz_file_name}'
sim_folder_scratch = f'/scratch/{bh5670}/{sim}/{rank}'
sim_folder_home = f'{sim}/{rank}' # sim folder at home exists. you create later {rank} folder
if not os.path.exists(sim_folder_scratch):
os.mkdir(sim_folder_scratch)
else:
rmtree(sim_folder_scratch
) # leftovers from previous simulations will be removed
os.mkdir(sim_folder_scratch) # and the new folder will be created
# xyz object created, normal xyz file is created at scratch
my_xyz_file_obj = XYZ.from_file(
xyz_file_location) # object created using the file from home
xyz_at_scratch = sim_folder_scratch + '/' + xyz_file_name #
my_xyz_file_obj.write(xyz_at_scratch) # writes a normal xyz (into scratch)
# my molecule object is created. It will serve as a DB record
my_new_mol = Cp2kOutput(rank)
# rel_cutoff: 40; cutoff: 300; abc = 10
my_abc = str(my_xyz_file_obj.compute_box_size(offset=my_offset))[1:-2]
# GLOBAL SETTINGS
#
## base settings ##
my_potential_file_name = 'POTENTIAL'
my_potential = 'ALL'
my_project_name = "this_is_template"
# my_ri_aux_basis_set = 'RI-5Z' # often fails
organic_elements = [
'H', 'C', 'N', 'O', 'F', 'P', 'S', 'Cl', 'Br', 'B', 'I'
]
my_elements = organic_elements
inp_file_name = 'test_2345.inp'
my_vdw_parameters_file = 'dftd3.dat'
activate_vdw = False
activate_outer_scf = False
wf_corr_num_proc = 1 # 16 in the ref paper; -1 to use all
########################################### CREATE TEMPLATE FOR TWO RUNS ###########################################
calc = CP2K()
calc.working_directory = './'
calc.project_name = 'artem_gw_project'
calc.mpi_n_processes = 1
# pycp2k objects
CP2K_INPUT = calc.CP2K_INPUT
FORCE_EVAL = CP2K_INPUT.FORCE_EVAL_add()
FORCE_EVAL.Method = 'QUICKSTEP'
SUBSYS = FORCE_EVAL.SUBSYS
DFT = FORCE_EVAL.DFT
XC = DFT.XC
SCF = DFT.SCF
OUTER_SCF = DFT.SCF.OUTER_SCF
####################################################################################################################
# GLOBAL #
# FORCE EVAL #
set_global(CP2K_INPUT, project_name=my_project_name)
## SUBSYS ##
set_unperiodic_cell(SUBSYS, abc=my_abc)
set_nonperiodic_poisson(DFT)
set_topology(SUBSYS, xyz_file_name=xyz_file_name)
center_coordinates(SUBSYS)
## END SUBSYS ##
## DFT ##
set_dft(DFT,
potential_file_name=my_potential_file_name,
basis_set_file_name=basis_set_file_name)
set_cutoff(DFT, cutoff=cutoff, rel_cutoff=rel_cutoff, ngrids=5)
set_scf(DFT, eps_scf=1.0E-9, max_scf=500, scf_guess='ATOMIC')
add_ot(SCF, stepsize=0.05)
#
# add_outer_scf(OUTER_SCF)
set_pbe(XC) # we start with pbe
# set_pbe0(XC) no pbe0 in the beginning
set_qs(DFT, eps_default=1.0E-10, eps_pgf_orb=np.sqrt(1.0E-10))
# print_mo(DFT.PRINT)
if activate_vdw:
add_vdw(XC, vdw_parameters_file=my_vdw_parameters_file)
## END DFT ##
######################################## END: CREATE TEMPLATE ##########################################################
######################################## BEGIN: RUN CP2K TWO TIMES #####################################################
suffix = ['2', '3', '4'] # cardinal numbers of the database
# begin: input_from_yaml
# cp2k_exe_path = '/home/artem/soft/cp2k/cp2k-7.1/exe/local/cp2k.popt'
cp2k_exe_path = '/home/ws/bh5670/cp2k/cp2k-7.1/exe/local/cp2k.popt'
my_run_type = 'mpi'
for i_bs, suffix in enumerate(suffix):
# bs
calc_ = deepcopy(calc)
#
CP2K_INPUT_ = calc_.CP2K_INPUT
FORCE_EVAL_ = CP2K_INPUT_.FORCE_EVAL_list[0]
SUBSYS_ = FORCE_EVAL_.SUBSYS
DFT_ = FORCE_EVAL_.DFT
XC_ = DFT_.XC
SCF_ = DFT_.SCF
OUTER_SCF_ = DFT_.SCF.OUTER_SCF
#
set_global(CP2K_INPUT_, project_name=suffix)
add_elements(SUBSYS_,
elements=my_elements,
basis=my_basis_sets[i_bs],
aux_basis=my_ri_basis_sets[i_bs],
pot=my_potential)
# bs
output_file = f'out_{suffix}.out'
ot_file_name = 'OT_' + f'{suffix}_' + inp_file_name
diag_file_name = 'DIAG_' + f'{suffix}_' + inp_file_name
# end: input
if i_bs != 0:
set_scf(DFT_, eps_scf=1.0E-8, max_scf=500, scf_guess='RESTART'
) # TZ,QZ will start from RESTART of the DZ,QZ
try:
copy(sim_folder_scratch + '/' + f'{int(suffix)-1}-RESTART.wfn',
sim_folder_scratch + '/' + f'{suffix}-RESTART.wfn')
print('copied restart file 2->3 or 3->4')
except:
print('not succesfull copy of the restart file')
elif i_bs == 0:
set_scf(DFT_, eps_scf=1.0E-8, max_scf=500,
scf_guess='ATOMIC') # DZ with ATOMIC guess
# OT run to converge quickly
calc_.write_input_file(sim_folder_scratch + '/' + ot_file_name)
# first run
print(f"Running PBE with OT (basis set = {suffix})...")
if not dummy_run:
cp2k_run(input_file=ot_file_name,
xyz_file=xyz_file_name,
run_type=my_run_type,
np=threads,
output_file=f'out_ot_{suffix}.out',
cp2k_executable=cp2k_exe_path,
execution_directory=sim_folder_scratch)
# end: first run
print(f"I have finished cp2k with OT (basis set = {suffix})")
# DIAGONALIZATION RUN to reliably compute H**O and then GW
# remove the OT method
remove_ot(SCF_)
# change calculations to a diagonalization
add_diagonalization(SCF_)
# add_smear(SCF_) # uses final T.
add_mixing(SCF_) # add or not?
add_mos(SCF_, added_mos=1000)
# plot h**o/lumo
#set_pbe0(XC_) # we want G0W0@PBE0. no pbe0 in the beginning
print_mo_cubes(DFT_.PRINT, nhomo=10,
nlumo=10) # all HOMOs are typicall plotted
set_scf(DFT_, eps_scf=1E-6, max_scf=200)
# add G0W0!
add_gw_ver_0(XC_,
ev_sc_iter=1,
wf_corr_num_proc=wf_corr_num_proc,
rpa_num_quad_points=100,
max_memory_wf=4000,
max_memory_hf=500,
corr_occ=1,
corr_virt=1) # GW!
# it is important to keep WF memory smaller than HF memory, otherwise, it crashes
calc_.write_input_file(sim_folder_scratch + '/' + diag_file_name)
# second run
print(f"Running G0W0 with DIAG (basis set = {suffix})...")
my_out_file2 = f'out_diag_{suffix}.out'
if not dummy_run:
cp2k_run(input_file=diag_file_name,
xyz_file=xyz_file_name,
output_file=my_out_file2,
run_type=my_run_type,
np=threads,
cp2k_executable=cp2k_exe_path,
execution_directory=sim_folder_scratch)
print(f"I have finished cp2k with DIAG (basis set = {suffix})")
# extract h**o/lumo and gw h**o/lumo from the cp2k output file:
path_to_out2_file = sim_folder_scratch + '/' + my_out_file2
# extract from the output
try:
num_orb = extract_number_of_independent_orbital_function(
path_to_out2_file)
print(
f'basis set = {suffix}, number of independent orbital functions: {num_orb}'
)
except:
print('number of orbatals was not extracted')
num_orb = 'not extracted'
try:
homos, lumos = [], []
homos, lumos = return_homo_lumo(path_to_out2_file)
print(f'basis set = {suffix} ', 'h**o = ',
homos[-1] * eV_to_Hartree(), ' eV')
print(f'basis set = {suffix} ', 'lumo = ',
lumos[0] * eV_to_Hartree(), ' eV')
h**o = homos[-1] * eV_to_Hartree()
lumo = lumos[0] * eV_to_Hartree()
except:
print(f'H**o/Lumo were not extracted')
h**o = 'not extracted'
lumo = 'not extracted'
try:
gw_occ, gw_vir, homo_, lumo_ = return_gw_energies(
path_to_out2_file)
if isinstance(h**o, str) and isinstance(lumo, str):
h**o = homo_
lumo = lumo_
print(f'basis set = {suffix} ', 'h**o = ', h**o, ' eV')
print(f'basis set = {suffix} ', 'lumo = ', lumo, ' eV')
print(f'basis set = {suffix} ', 'gw h**o = ', gw_occ, ' eV')
print(f'basis set = {suffix} ', 'gw lumo = ', gw_vir, ' eV')
except:
print("GW energies were not extracted")
gw_occ = 'not extracted'
gw_vir = 'not extracted'
del calc_
# put computed data into the molecule object
my_new_mol.add_energies(int(suffix), h**o, lumo, gw_occ, gw_vir)
my_new_mol.add_num_orbitals(int(suffix), num_orb)
my_new_mol.extrapolate_energy()
db_record = my_new_mol.yield_dict(
) # this dict will be written into yaml. it will be a record in the global library
#
print("\nI am done\n")
print('saving to DB...')
with open(f'{database_folder}/DB_{rank}.yaml', 'w') as stream:
yaml.safe_dump(db_record, stream)
print(f"saved to {database_folder}/DB_{rank}.yaml")
print('I will remove the content the sim folder')
# Clean up before leave
status = my_new_mol.status()
if status == 'all_extracted': # all quantities are extracted
if debug:
print(
f'status: {status}, but debug is on ==> will move {sim_folder_scratch} to {sim_folder_home}'
)
copytree(sim_folder_scratch,
sim_folder_home) # will rewrite the folder
else:
print(f'status: {status} ==> will remove {sim_folder_scratch}')
try_to_remove_folder(sim_folder_scratch)
else:
print(f'status: {status} ==> will copy failed sim folder from scratch')
#if not os.path.exists(sim_folder_home):
#os.mkdir(sim_folder_home) # will overwrite if exists
copytree(sim_folder_scratch,
sim_folder_home) # will rewrite the folder
print(f"I have copied {sim_folder_scratch} to {sim_folder_home}")
try_to_remove_folder(sim_folder_scratch)
def main():
try:
scratch = os.environ[
'SCRATCH'] # SCRATCH has to be in the env var dict. Normally, it is.
except:
scratch = 'scratch'
# parser begin
parser = argparse.ArgumentParser(description='rank and num of cpus')
parser.add_argument('-rank') # array job number
parser.add_argument(
'-num_cpus') # number of cpus you request for every array job
parser.add_argument('-i') # input_from_yaml yaml file
parser.add_argument(
'-mol_ids'
) # mol_ids to simulate (without prefix and suffix) todo: not used?
args = parser.parse_args()
# parser end
# yaml file
yaml_file_name = args.i
with open(yaml_file_name) as stream:
input = yaml.load(stream=stream)
# end: yaml file
# todo: think over because it is imported twice
# end: run-or-check settings
debug = input['debug']
dummy_run = input['dummy_run']
# if not at cluster: test
# debug = True
# dummy_run = True
# end: if not at cluster
# folders names
sim = input['folder_names']['simulations']
db = input['folder_names']['database']
bh5670 = input['folder_names'][
'scratch'] # the outermost folder in the scratch folder where all other data are put
prefix_xyz_file_name = input['prefix_xyz_file_name']
my_offset = input['molecule_vacuum_offset']
try:
mpi = input['mpi']
except:
mpi = 'openmpi'
# parsing input_from_yaml
threads = int(
args.num_cpus
) # cpus used to compute. I do not subtract 1. This does not help
# mol_id = parse_mixed_list()
path_to_mol_ids_default = 'db/trash_db_numbers.csv' # simulate mols that did not fully converged
# path_to_mol_ids_default = 'db/missing_num.csv' # simulate mols that are missing in the range of the simulated mols
# by default: missing_numbers
try:
path_to_mol_ids = args.mol_ids
if path_to_mol_ids is None:
raise TypeError
except TypeError:
path_to_mol_ids = path_to_mol_ids_default
finally:
with open(path_to_mol_ids, 'r') as stream:
csv_reader = csv.reader(stream)
all_numbers = csv_reader.__next__(
) # only one line in this csv format file, so we do not loop over
rank = all_numbers[int(args.rank) - 1]
rank = '{:0>6}'.format(
rank
) # transform rank from '1' to '000001' format. This is not a general thing
xyz_file_name = f'{prefix_xyz_file_name}_{rank}.xyz'
xyz_file_location = f'{prefix_xyz_file_name}/{xyz_file_name}'
db_record_path = f'{db}/DB_{rank}.yaml' # file where the results will be saved todo: raeum es alles auf!
# check is the output exists
if os.path.exists(db_record_path):
print(
f'The simulation results of mol. {rank} is already in the folder of reference'
)
exit()
# end: check if the output exists
if not dummy_run:
sim_folder_scratch = f'/{scratch}/{bh5670}/{sim}/{rank}'
else:
sim_folder_scratch = f'{scratch}/{bh5670}/{sim}/{rank}'
sim_folder_home = f'{sim}/{rank}' # sim folder at home exists. you create later {rank} folder
if not os.path.exists(sim_folder_scratch):
os.mkdir(sim_folder_scratch)
else:
rmtree(sim_folder_scratch
) # leftovers from previous simulations will be removed
os.mkdir(sim_folder_scratch) # and the new folder will be created
# xyz object created, normal xyz file is created at scratch
try:
my_xyz_file_obj = XYZ.from_file(
xyz_file_location) # object created using the file from home
except: # test
my_xyz_file_obj = XYZ.from_file(
'H2O.xyz') # object created using the file from home
xyz_at_scratch = sim_folder_scratch + '/' + xyz_file_name #
my_xyz_file_obj.write(xyz_at_scratch) # writes a normal xyz (into scratch)
# my molecule object is created. It will serve as a DB record
my_new_mol = Cp2kOutput(rank)
# rel_cutoff: 40; cutoff: 300; abc = 10
my_abc = str(my_xyz_file_obj.compute_box_size(offset=my_offset))[1:-2]
input['my_abc'] = my_abc
input['xyz_file_name'] = xyz_file_name
# misc
wf_corr_num_proc = 0 # 16 in the ref paper; -1 to use all
inp_file_name = 'test_2344.inp' # base file name
######################################## BEGIN: RUN CP2K TWO TIMES #####################################################
# suffix = ['2', '3', '4'] # cardinal numbers of the database
# begin: input_from_yaml
# cp2k_exe_path = '/home/artem/soft/cp2k/cp2k-7.1/exe/local/cp2k.popt'
#cp2k_exe_path = '/home/ws/bh5670/cp2k/cp2k-7.1/exe/local/cp2k.popt'
cp2k_exe_path = input['cp2k_exe_path']
my_run_type = 'mpi'
suffix = input[
'basis_set_suffix'] # todo: fix DZ --> 2, TZ --> 3, QZ --> 4
# this will initialize class variables (that is the class) according to the input
# actually, this is probably a bad idea to make it like that, because if one forgets doing so,
# class functions will not work
InputFactory.set_constants(input_from_yaml=input)
# --> my_cp2k_run: condensed function with just 2 argument.
# my_inp_file, my_out_file: return names
# reason: its other parameters are the same for all 6 (or more runs)
# this is nothing more than a shorthand, this is why it is ugly
def my_inp_file(suf, ot_or_diag):
return f'{ot_or_diag}_{suf}.inp'
def my_out_file(suf, ot_or_diag):
return f'out_{ot_or_diag}_{suf}.out'
def my_cp2k_run(suf='2', ot_or_diag='ot'):
cp2k_run(input_file=my_inp_file(suf, ot_or_diag),
output_file=my_out_file(suf, ot_or_diag),
xyz_file=xyz_file_name,
run_type=my_run_type,
np=threads,
cp2k_executable=cp2k_exe_path,
execution_directory=sim_folder_scratch,
type_mpi=mpi)
# <-- my_cp2k_run
# ot_file_names = [f'OT_{suffix}_{inp_file_name}' for suffix in suffix]
# diag_file_names =[f'DIAG_{suffix}_{inp_file_name}' for suffix in suffix]
# out_ot_file_names = [f'out_ot_{suffix}.out' for suffix in suffix]
# out_diag_file_names =[f'out_diag_{suffix}.out' for suffix in suffix]
print('I am HERE')
for i_bs, suf in enumerate(suffix):
# --> OT dft. (OT = orbital transformation)
dft_ot_simulation = InputFactory.new_dft_ot(i_bs)
dft_ot_simulation.write_input_file(
f"{sim_folder_scratch}/{my_inp_file(suf=suf, ot_or_diag='ot')}")
# OT dft run below ...
# ... but before, we copy the RESTART from the previous basis set (it exists unless for the smallest basis set)
try_to_copy_previous_restart_file(
i_bs=i_bs, sim_folder_scratch=sim_folder_scratch, suf=suf)
print(f"Running PBE with OT (basis set = {suf})...")
if not dummy_run:
my_cp2k_run(suf=suf, ot_or_diag='ot')
print(f"I have finished cp2k with OT (basis set = {suf})")
# <-- OT dft
# --> GW following DIAG dft. (DIAG = diagonalization)
diag_out_file = f"{sim_folder_scratch}/{my_out_file(suf=suf, ot_or_diag='diag')}"
diag_inp_file = f"{sim_folder_scratch}/{my_inp_file(suf=suf, ot_or_diag='diag')}"
gw_diag_simulations = InputFactory.new_gw(i_bs)
gw_diag_simulations.write_input_file(diag_inp_file)
print(f"Running G0W0 with DIAG (basis set = {suf})...")
if not dummy_run:
my_cp2k_run(suf=suf, ot_or_diag='diag')
print(f"I have finished cp2k with DIAG (basis set = {suf})")
# --> extract (from diag out)
# extract number of orbitals:
try:
num_orb = extract_number_of_independent_orbital_function(
diag_out_file)
print(
f'basis set = {suf}, number of independent orbital functions: {num_orb}'
)
except:
print('number of orbitals was not extracted')
num_orb = 'not extracted'
# extract energies:
try:
homos, lumos = [], []
homos, lumos = return_homo_lumo(diag_out_file)
print(f'basis set = {suffix} ', 'h**o = ',
homos[-1] * eV_to_Hartree(), ' eV')
print(f'basis set = {suffix} ', 'lumo = ',
lumos[0] * eV_to_Hartree(), ' eV')
h**o = homos[-1] * eV_to_Hartree()
lumo = lumos[0] * eV_to_Hartree()
except:
print(f'H**o/Lumo were not extracted')
h**o = 'not extracted'
lumo = 'not extracted'
try:
gw_occ, gw_vir, homo_, lumo_ = return_gw_energies(
diag_out_file)
h**o, lumo = redefine_homo_lumo_if_not_extracted_before(
homo_, lumo_, h**o, lumo)
print_extracted_energies(suf, h**o, lumo, gw_occ,
gw_vir) # on a screen
except SCQPSolutionNotFound: # we know how to handle this error
print(
"GW is not extracted, because SCQPSolutionNotFound. Calling fallback ..."
)
# --> of the solution not found, it could be that the number of quad points is insufficent
gw_diag_simulations.CP2K_INPUT.FORCE_EVAL_list[
0].DFT.XC.WF_CORRELATION_list[
0].RI_RPA.Rpa_num_quad_points = 500
print(
"I write the fallback input file where num of q points = 500. It has the same name as before?"
)
gw_diag_simulations.write_input_file(diag_inp_file)
my_cp2k_run(suf=suf, ot_or_diag='diag')
except SCFNotConvergedNotPossibleToRunMP2:
print(
"GW is not extracted, because SCFNotConvergedNotPossibleToRunMP2. Calling fallback ..."
)
print('NOT IMPLEMENTED')
finally:
try:
gw_occ, gw_vir, homo_, lumo_ = return_gw_energies(
diag_out_file)
h**o, lumo = redefine_homo_lumo_if_not_extracted_before(
homo_, lumo_, h**o, lumo)
print_extracted_energies(suffix, h**o, lumo, gw_occ,
gw_vir) # on a screen
# <---
except:
print(
"GW energies were not extracted even in the fallback")
gw_occ = 'not extracted'
gw_vir = 'not extracted'
del dft_ot_simulation, gw_diag_simulations
# put computed data into the molecule object
my_new_mol.add_energies(int(suf), h**o, lumo, gw_occ, gw_vir)
my_new_mol.add_num_orbitals(int(suf), num_orb)
my_new_mol.extrapolate_energy() # level up?
db_record = my_new_mol.yield_dict(
) # this dict will be written into yaml. it will be a record in the global library
# <-- EMD: GW run and extraction
####################################### END: RUN CP2K TWO TIMES #####################################################
print("\nI am done\n")
if not dummy_run:
print('saving to DB...')
with open(f'{db}/DB_{rank}.yaml', 'w') as stream:
yaml.safe_dump(db_record, stream)
print(f"saved to {db}/DB_{rank}.yaml")
print('I will remove the content of the sim folder')
# Clean up before leave
status = my_new_mol.status()
if status == 'all_extracted': # all quantities are extracted
if debug:
print(
f'status: {status}, but debug is on ==> will move {sim_folder_scratch} to {sim_folder_home}'
)
copytree(sim_folder_scratch, sim_folder_home,
dirs_exist_ok=True) # will rewrite the folder
else:
print(f'status: {status} ==> will remove {sim_folder_scratch}')
try_to_remove_folder(sim_folder_scratch)
else:
print(f'status: {status} ==> will copy failed sim folder from scratch')
#if not os.path.exists(sim_folder_home):
#os.mkdir(sim_folder_home) # will overwrite if exists
try:
copytree(sim_folder_scratch,
sim_folder_home) # will rewrite the folder?
print(f"I have copied {sim_folder_scratch} to {sim_folder_home}")
except:
print(
f"I could not copy {sim_folder_scratch} to {sim_folder_home}")
try_to_remove_folder(sim_folder_scratch)
def main():
def my_out_file(suf, ot_or_diag):
return f'out_{ot_or_diag}_{suf}.out'
SUFFIX = ['2', '3']
my_path = os.path.abspath('')
print(my_path)
all_folders = os.listdir('sim')
print(f'all folders of interest = {all_folders}')
path_to_folders = [
os.path.abspath('sim' + '/' + folder) for folder in all_folders
]
print(path_to_folders)
for folder, path in zip(all_folders, path_to_folders):
db_record_path = f'db/DB_{folder}.yaml'
if os.path.exists(db_record_path):
print(
f'The simulation results of mol. {folder} is already in the folder of reference'
)
exit()
print(f'\nTHIS IS MOL_NUM {folder}\n')
my_new_mol = Cp2kOutput(folder)
for suf in SUFFIX:
diag_out_file = f"sim/{folder}/{my_out_file(suf=suf, ot_or_diag='diag')}"
#
try:
num_orb = extract_number_of_independent_orbital_function(
diag_out_file)
print(
f'basis set = {suf}, number of independent orbital functions: {num_orb}'
)
except:
print('number of orbitals was not extracted')
num_orb = 'not extracted'
# extract energies:
try:
homos, lumos = [], []
homos, lumos = return_homo_lumo(diag_out_file)
print(f'basis set = {suf} ', 'h**o = ',
homos[-1] * eV_to_Hartree(), ' eV')
print(f'basis set = {suf} ', 'lumo = ',
lumos[0] * eV_to_Hartree(), ' eV')
h**o = homos[-1] * eV_to_Hartree()
lumo = lumos[0] * eV_to_Hartree()
except:
print(f'H**o/Lumo were not extracted')
h**o = 'not extracted'
lumo = 'not extracted'
try:
gw_occ, gw_vir, homo_, lumo_ = return_gw_energies(
diag_out_file)
h**o, lumo = redefine_homo_lumo_if_not_extracted_before(
homo_, lumo_, h**o, lumo)
print_extracted_energies(suf, h**o, lumo, gw_occ,
gw_vir) # on a screen
except SCFNotConvergedNotPossibleToRunMP2:
print(
"GW is not extracted, because SCFNotConvergedNotPossibleToRunMP2. Calling fallback ..."
)
print('NOT IMPLEMENTED')
finally:
try:
gw_occ, gw_vir, homo_, lumo_ = return_gw_energies(
diag_out_file)
h**o, lumo = redefine_homo_lumo_if_not_extracted_before(
homo_, lumo_, h**o, lumo)
print_extracted_energies(suf, h**o, lumo, gw_occ,
gw_vir) # on a screen
# <---
except:
print(
"GW energies were not extracted even in the fallback")
gw_occ = 'not extracted'
gw_vir = 'not extracted'
# put computed data into the molecule object
my_new_mol.add_energies(int(suf), h**o, lumo, gw_occ, gw_vir)
my_new_mol.add_num_orbitals(int(suf), num_orb)
my_new_mol.extrapolate_energy() # level up?
db_record = my_new_mol.yield_dict(
) # this dict will be written into yaml. it will be a record in the global library
# <-- EMD: GW run and extraction
print('saving to DB...')
with open(f'db/DB_{folder}.yaml', 'w') as stream:
yaml.safe_dump(db_record, stream)
print(f"saved to db/DB_{folder}.yaml")
#
print('done')
def main():
try:
scratch = os.environ['SCRATCH'] # SCRATCH has to be in the env var dict. Normally, it is.
except:
scratch = 'scratch'
# parser begin
parser = argparse.ArgumentParser(description='rank and num of cpus')
parser.add_argument('-rank') # array job number
parser.add_argument('-num_cpus') # number of cpus you request for every array job
parser.add_argument('-i') # input_from_yaml yaml file
parser.add_argument('-mol_ids') # path to csv file with some identifiers of molecules to be simulated.
# Identifiers may be: numbers of the molecules (6 digits) or the full names of the molecules mol_ids to simulate
args = parser.parse_args()
# parser end
# yaml file. my_input is the dictionaty containing input information.
yaml_file_name = args.i
with open(yaml_file_name) as stream:
my_input = yaml.load(stream=stream, Loader=yaml.SafeLoader)
# end: yaml file
# todo: think over because it is imported twice
# end: run-or-check settings
debug = my_input['debug'] # if True will copy the content of the scratch folder back to sim/* folder. Actually, should be used
dummy_run = my_input['dummy_run']
# if not at cluster: test
# debug = True
# dummy_run = True
# end: if not at cluster
# folders names
sim = my_input['folder_names']['simulations']
db = my_input['folder_names']['database']
bh5670 = my_input['folder_names']['scratch'] # the outermost folder in the scratch folder where all other data are put
my_offset = my_input['molecule_vacuum_offset']
try:
mpi = my_input['mpi']
except:
mpi = 'openmpi'
# parsing input_from_yaml
threads = int(args.num_cpus) # cpus used to compute. I do not subtract 1. This does not help
# mol_id = parse_mixed_list()
def determine_file_name_path(my_input, args):
"""
Depending on the format of the input dictionary (my_input['db_format']),
returns the name of the input file and the path to it
@param my_input: input dictionary, which was read in from the yaml input file
@param args: command line arguments list
@return: xyz_file_name, xyz_file_path, mol_identifier (self-explained)
"""
try:
db_format = my_input['db_format'] # 'general' means that names are full mols identifiers
except KeyError:
db_format = 'dsgdb9nsd' # this is the format of the corresponding dataset only. default
try:
path_to_mol_ids = args.mol_ids
except TypeError:
print(f'path_to_mol_ids {args.mol_ids} is not found. Exiting...')
exit()
with open(path_to_mol_ids, 'r') as stream:
csv_reader = csv.reader(stream)
all_mols_ids = csv_reader.__next__() # only one line in this csv format file, so we do not loop over
if db_format =='dsgdb9nsd':
prefix_xyz_file_name = my_input['prefix_xyz_file_name']
mol_identifier_variable_digits = all_mols_ids[int(args.rank) - 1] # the variable "rank" is not actually a rank. This is here 6 digits read from *.csv file (only for db_format = 'gdb...' or if it is nnot specified).
mol_identifier_6_digits = '{:0>6}'.format(mol_identifier_variable_digits) # transform rank from '1' to '000001' format. This is not a general thing
xyz_file_name = f'{prefix_xyz_file_name}_{mol_identifier_6_digits}.xyz' # only file name
xyz_file_path = f'../{prefix_xyz_file_name}/{xyz_file_name}' # path to the file. db outside working folder
return xyz_file_name, xyz_file_path, mol_identifier_6_digits
elif db_format == 'general':
try:
prefix_xyz_file_name = my_input['prefix_xyz_file_name'] # here, xyz file name is the name of the db folder.
except KeyError:
prefix_xyz_file_name = my_input['dataset_name']
mol_identifier = all_mols_ids[int(args.rank) - 1].split('.')[0]
xyz_file_name = f'{mol_identifier}.xyz' # only file name
xyz_file_path = f'../{prefix_xyz_file_name}/{xyz_file_name}' # path to the file. db outside working folder
return xyz_file_name, xyz_file_path, mol_identifier
else:
print('Unknown db_format. Exiting...')
exit()
xyz_file_name, xyz_file_path, mol_identifier = determine_file_name_path(my_input, args)
# db -->
# check is the output exists in 'db' folder
db_record_path = f'{db}/DB_{mol_identifier}.yaml' # file where the results will be saved
if os.path.exists(db_record_path):
print(f'The simulation results of mol. {mol_identifier} is already in the folder of reference')
exit()
# here one can check if the DB_ file is not broken
# end: check if the output exists
# <-- db
# scratch -->
# this below makes something. look carefully!
if not dummy_run:
print('This a productive run')
sim_folder_scratch = f'{scratch}/{bh5670}/{sim}/{mol_identifier}'
print(f'I set a sim_folder_scratch to: {sim_folder_scratch}')
else:
print('This is dummy run')
sim_folder_scratch = f'{scratch}/{bh5670}/{sim}/{mol_identifier}'
print(f'I set a sim_folder_scratch to: {sim_folder_scratch}')
os.makedirs(sim_folder_scratch, exist_ok=True)
#<-- scratch
# sim -->
sim_folder_home = f'{sim}/{mol_identifier}' # sim folder at home exists (has to exist beforehand). you create later {mol_id} folder
if not os.path.exists(sim_folder_home): # home
os.mkdir(sim_folder_home)
else:
print(f"I have found the folder {sim_folder_home} in the sim folder and will try to copy it to scratch ....")
copytree(sim_folder_home, sim_folder_scratch, dirs_exist_ok=True) # will rewrite the folder
print('...done!')
print(f'now I will remove the {sim_folder_home} folder at home and create a new empty folder at its place...')
rmtree(sim_folder_home) # leftovers from previous simulations will be removed
os.mkdir(sim_folder_home) # and the new folder will be created
print('...done')
# <-- sim
# xyz object created, normal xyz file is created at scratch
# todo: H20.xyz is not relevant anymore
try:
my_xyz_file_obj = XYZ.from_file(xyz_file_path) # object created using the file from home
except FileNotFoundError: # test
my_xyz_file_obj = XYZ.from_file('H2O.xyz') # object created using the file from home
xyz_at_scratch = sim_folder_scratch + '/' + xyz_file_name #
my_xyz_file_obj.write(xyz_at_scratch) # writes a normal xyz (into scratch)
# my molecule object is created. It will serve as a DB record
my_new_mol = Cp2kOutput(mol_identifier)
# rel_cutoff: 40; cutoff: 300; abc = 10
my_abc = str(my_xyz_file_obj.compute_box_size(offset=my_offset))[1:-2]
# last changes to my_input: offsets and xyz_file_name
my_input['my_abc'] = my_abc
my_input['xyz_file_name'] = xyz_file_name
my_xyz_file_obj.identify_atom_types()
my_input['elements'] = my_xyz_file_obj.unique_atom_types # elements identified automatically
# misc
wf_corr_num_proc = 0 # 16 in the ref paper; -1 to use all
inp_file_name = 'test_2344.inp' # base file name
######################################## BEGIN: RUN CP2K TWO TIMES #####################################################
# suffix = ['2', '3', '4'] # cardinal numbers of the database
# begin: input_from_yaml
# cp2k_exe_path = '/home/artem/soft/cp2k/cp2k-7.1/exe/local/cp2k.popt'
# cp2k_exe_path = '/home/ws/bh5670/cp2k/cp2k-7.1/exe/local/cp2k.popt'
cp2k_exe_path = my_input['cp2k_exe_path']
my_run_type = 'mpi'
suffix = my_input['basis_set_suffix'] # todo: fix DZ --> 2, TZ --> 3, QZ --> 4
# this will initialize class variables (that is the class) according to the input
# actually, this is probably a bad idea to make it like that, because if one forgets doing so,
# class functions will not work
InputFactory.set_constants(input_from_yaml=my_input)
# --> my_cp2k_run: condensed function with just 2 argument.
# my_inp_file, my_out_file: return names
# reason: its other parameters are the same for all 6 (or more runs)
# this is nothing more than a shorthand, this is why it is ugly
def my_inp_file(suf, ot_or_diag):
return f'{ot_or_diag}_{suf}.inp'
def my_out_file(suf, ot_or_diag):
return f'out_{ot_or_diag}_{suf}.out'
def my_cp2k_run(suf='2', ot_or_diag='ot'):
cp2k_run(input_file=my_inp_file(suf, ot_or_diag),
output_file=my_out_file(suf, ot_or_diag),
xyz_file=xyz_file_name,
run_type=my_run_type,
np=threads,
cp2k_executable=cp2k_exe_path,
execution_directory=sim_folder_scratch,
type_mpi=mpi)
# <-- my_cp2k_run
# ot_file_names = [f'OT_{suffix}_{inp_file_name}' for suffix in suffix]
# diag_file_names =[f'DIAG_{suffix}_{inp_file_name}' for suffix in suffix]
# out_ot_file_names = [f'out_ot_{suffix}.out' for suffix in suffix]
# out_diag_file_names =[f'out_diag_{suffix}.out' for suffix in suffix]
print('I am HERE')
for i_bs, suf in enumerate(suffix):
# --> OT dft. (OT = orbital transformation)
dft_ot_simulation = InputFactory.new_dft_ot(i_bs)
ot_inp_file = f"{sim_folder_scratch}/{my_inp_file(suf=suf, ot_or_diag='ot')}"
dft_ot_simulation.write_input_file(ot_inp_file)
# OT dft run below ...
# ... but before, we copy the RESTART from the previous basis set (it exists unless for the smallest basis set)
try_to_copy_previous_restart_file(i_bs=i_bs, sim_folder_scratch=sim_folder_scratch, suf=suf)
print(f"Running PBE with OT (basis set = {suf})...")
if not dummy_run:
my_cp2k_run(suf=suf, ot_or_diag='ot')
print(f"I have finished cp2k with OT (basis set = {suf})")
# <-- OT dft
# --> GW following DIAG dft. (DIAG = diagonalization)
diag_out_file = f"{sim_folder_scratch}/{my_out_file(suf=suf, ot_or_diag='diag')}"
diag_inp_file = f"{sim_folder_scratch}/{my_inp_file(suf=suf, ot_or_diag='diag')}"
gw_diag_simulations = InputFactory.new_gw(i_bs)
gw_diag_simulations.write_input_file(diag_inp_file)
print(f"Running G0W0 with DIAG (basis set = {suf})...")
if not dummy_run:
my_cp2k_run(suf=suf, ot_or_diag='diag')
print(f"I have finished cp2k with DIAG (basis set = {suf})")
# --> extract (from diag out)
# extract number of orbitals:
try:
num_orb = extract_number_of_independent_orbital_function(diag_out_file)
print(f'basis set = {suf}, number of independent orbital functions: {num_orb}')
except:
print('number of orbitals was not extracted')
num_orb = 'not extracted'
# extract energies:
try:
homos, lumos = [], []
homos, lumos = return_homo_lumo(diag_out_file)
print(f'basis set = {suffix} ', 'h**o = ', homos[-1] * eV_to_Hartree(), ' eV')
print(f'basis set = {suffix} ', 'lumo = ', lumos[0] * eV_to_Hartree(), ' eV')
h**o = homos[-1] * eV_to_Hartree()
lumo = lumos[0] * eV_to_Hartree()
except:
print(f'H**o/Lumo were not extracted')
h**o = 'not extracted'
lumo = 'not extracted'
try: # first try to return gw energies -->
occ, vir, homo_, lumo_, occ_scf, vir_scf, occ_0, vir_0 = return_gw_energies_advanced(diag_out_file)
h**o, lumo = redefine_homo_lumo_if_not_extracted_before(homo_, lumo_, h**o, lumo)
print_extracted_energies(suf, h**o, lumo, occ, vir) # on a screen
except (IterationLimit, LargeSigc, SCQPSolutionNotFound, NaNInGW): # 20 iterations
try: # xyz + 10
print("GW is extracted, but scf is not converged, because of IterationLimit. Calling fallback ...")
my_abc_plus_10 = str(my_xyz_file_obj.compute_box_size(offset=my_offset+10.0))[1:-2] # todo: hard
# replay ot with a larger xyz space +10
# ot
dft_ot_simulation.CP2K_INPUT.FORCE_EVAL_list[0].SUBSYS.CELL.Abc = my_abc_plus_10
dft_ot_simulation.write_input_file(ot_inp_file)
print("Replay ot with xyz + 10")
my_cp2k_run(suf=suf, ot_or_diag='ot')
print("... ot succesfull")
# diag
print("diag Replay ot with xyz + 10 and Femi offset of 5E-2")
gw_diag_simulations.CP2K_INPUT.FORCE_EVAL_list[0].SUBSYS.CELL.Abc = my_abc_plus_10
gw_diag_simulations.CP2K_INPUT.FORCE_EVAL_list[0].DFT.XC.WF_CORRELATION_list[0].RI_RPA.RI_G0W0.Fermi_level_offset = 4.0E-2
gw_diag_simulations.write_input_file(diag_inp_file)
my_cp2k_run(suf=suf, ot_or_diag='diag')
print("... diag succesfull")
# the following section is necessary to catch the error:
occ, vir, homo_, lumo_, occ_scf, vir_scf, occ_0, vir_0 = return_gw_energies_advanced(diag_out_file)
h**o, lumo = redefine_homo_lumo_if_not_extracted_before(homo_, lumo_, h**o, lumo)
print_extracted_energies(suf, h**o, lumo, occ, vir) # on a screen
except (IterationLimit, LargeSigc, SCQPSolutionNotFound, NaNInGW):
try:
print("GW is extracted, but scf is not converged AGAIN, because of IterationLimit. Calling fallback ...")
# diag 200 Q points
gw_diag_simulations.CP2K_INPUT.FORCE_EVAL_list[0].DFT.XC.WF_CORRELATION_list[0].RI_RPA.Rpa_num_quad_points = 200 # this should help as well
gw_diag_simulations.write_input_file(diag_inp_file)
my_cp2k_run(suf=suf, ot_or_diag='diag')
print("I write the fallback input file with QUAD points = 200")
gw_diag_simulations.write_input_file(diag_inp_file)
my_cp2k_run(suf=suf, ot_or_diag='diag')
print("... diag succesful")
# the following section is necessary to catch the error:
occ, vir, homo_, lumo_, occ_scf, vir_scf, occ_0, vir_0 = return_gw_energies_advanced(diag_out_file)
h**o, lumo = redefine_homo_lumo_if_not_extracted_before(homo_, lumo_, h**o, lumo)
print_extracted_energies(suf, h**o, lumo, occ, vir) # on a screen
except (IterationLimit, LargeSigc, SCQPSolutionNotFound):
# replay ot with a larger cutoff then make diag with a larger cutoff
# ot
print("GW is extracted, but scf is not converged AGAIN AGAIN, because of IterationLimit. Calling fallback ...")
print("Replay diag with FERMI offset 10E-2!")
gw_diag_simulations.CP2K_INPUT.FORCE_EVAL_list[0].DFT.XC.WF_CORRELATION_list[0].RI_RPA.RI_G0W0.Fermi_level_offset = 10.0E-2
gw_diag_simulations.write_input_file(diag_inp_file)
my_cp2k_run(suf=suf, ot_or_diag='diag')
print("... diag succesful")
except SCQPSolutionNotFound: # we know how to handle this error
try:
print("GW is not extracted, because SCQPSolutionNotFound. Calling fallback ...")
# --> of the solution not found, it could be that the number of quad points is insufficent
gw_diag_simulations.CP2K_INPUT.FORCE_EVAL_list[0].DFT.XC.WF_CORRELATION_list[0].RI_RPA.RI_G0W0.Crossing_search = 'BISECTION' # this alone does not always work
print("I write the fallback input file the crossing search is set to BISECTION")
gw_diag_simulations.write_input_file(diag_inp_file)
my_cp2k_run(suf=suf, ot_or_diag='diag')
# the following section is necessary to catch the error:
occ, vir, homo_, lumo_, occ_scf, vir_scf, occ_0, vir_0 = return_gw_energies_advanced(diag_out_file)
h**o, lumo = redefine_homo_lumo_if_not_extracted_before(homo_, lumo_, h**o, lumo)
print_extracted_energies(suf, h**o, lumo, occ, vir) # on a screen
except SCQPSolutionNotFound:
print("GW is not extracted, because SCQPSolutionNotFound. Calling second fallback ...")
gw_diag_simulations.CP2K_INPUT.FORCE_EVAL_list[0].DFT.XC.WF_CORRELATION_list[0].RI_RPA.Rpa_num_quad_points = 500 # this should help as well
gw_diag_simulations.CP2K_INPUT.FORCE_EVAL_list[0].DFT.XC.WF_CORRELATION_list[0].RI_RPA.RI_G0W0.Crossing_search = 'BISECTION'
print("I write the fallback input file with QUAD points = 500")
gw_diag_simulations.write_input_file(diag_inp_file)
my_cp2k_run(suf=suf, ot_or_diag='diag')
except SCFNotConvergedNotPossibleToRunMP2:
print("GW is not extracted, because SCFNotConvergedNotPossibleToRunMP2. Calling fallback ...")
# replay ot with a larger cutoff then make diag with a larger cutoff
# ot
dft_ot_simulation.CP2K_INPUT.FORCE_EVAL_list[0].DFT.MGRID.Cutoff = 1000
dft_ot_simulation.CP2K_INPUT.FORCE_EVAL_list[0].DFT.MGRID.Rel_cutoff = 100
dft_ot_simulation.write_input_file(ot_inp_file)
print("Replay ot with cutoff of 100 rel_cutoff of 100...")
my_cp2k_run(suf=suf, ot_or_diag='ot')
print("... ot succesfull")
# diag
gw_diag_simulations.CP2K_INPUT.FORCE_EVAL_list[0].DFT.MGRID.Cutoff = 1000
gw_diag_simulations.CP2K_INPUT.FORCE_EVAL_list[0].DFT.MGRID.Rel_cutoff = 100
gw_diag_simulations.write_input_file(diag_inp_file)
my_cp2k_run(suf=suf, ot_or_diag='diag')
# print('NOT IMPLEMENTED')
except NaNInGW:
try:
print("GW is not extracted, because there is a NaN in the last frame of the SCF loop. Calling fallback")
gw_diag_simulations.CP2K_INPUT.FORCE_EVAL_list[0].DFT.XC.WF_CORRELATION_list[0].RI_RPA.RI_G0W0.Crossing_search = 'BISECTION'
print("I wrote the fallback. The crossing search is set to BISECTION")
gw_diag_simulations.write_input_file(diag_inp_file)
my_cp2k_run(suf=suf, ot_or_diag='diag')
# print("NOT IMPLEMENTED")
occ, vir, homo_, lumo_, occ_scf, vir_scf, occ_0, vir_0 = return_gw_energies_advanced(diag_out_file)
h**o, lumo = redefine_homo_lumo_if_not_extracted_before(homo_, lumo_, h**o, lumo)
print_extracted_energies(suf, h**o, lumo, occ, vir) # on a screen
except (NaNInGW, SCQPSolutionNotFound):
print("GW is not extracted, because NaNInGW AGAIN. Calling second fallback (BISECTION and num_quad_points = 500) ...")
gw_diag_simulations.CP2K_INPUT.FORCE_EVAL_list[0].DFT.XC.WF_CORRELATION_list[0].RI_RPA.Rpa_num_quad_points = 500 # this should help as well
gw_diag_simulations.CP2K_INPUT.FORCE_EVAL_list[0].DFT.XC.WF_CORRELATION_list[0].RI_RPA.RI_G0W0.Crossing_search = 'BISECTION'
print("I wrote the fallback input file with QUAD points = 500")
gw_diag_simulations.write_input_file(diag_inp_file)
my_cp2k_run(suf=suf, ot_or_diag='diag')
finally:
try:
occ, vir, homo_, lumo_, occ_scf, vir_scf, occ_0, vir_0 = return_gw_energies_advanced(diag_out_file)
h**o, lumo = redefine_homo_lumo_if_not_extracted_before(homo_, lumo_, h**o, lumo)
print_extracted_energies(suf, h**o, lumo, occ, vir) # on a screen
# <---
except:
print("GW energies were not extracted even in the fallback")
# occ = 'not extracted'
# vir = 'not extracted'
occ, vir, occ_scf, vir_scf, occ_0, vir_0 = ['not extracted']*6
del dft_ot_simulation, gw_diag_simulations
# put computed data into the molecule object
my_new_mol.add_energies_advanced(int(suf), h**o, lumo, occ, vir, occ_0, vir_0, occ_scf, vir_scf)
my_new_mol.add_num_orbitals(int(suf), num_orb)
my_new_mol.extrapolate_energy_advanced() # level up?
db_record = my_new_mol.yield_dict() # this dict will be written into yaml. it will be a record in the global library
# <-- EMD: GW run and extraction
####################################### END: RUN CP2K TWO TIMES #####################################################
print("\nI am done\n")
if not dummy_run:
print('saving to DB...')
with open(f'{db}/DB_{mol_identifier}.yaml', 'w') as stream:
yaml.safe_dump(db_record, stream)
print(f"saved to {db}/DB_{mol_identifier}.yaml")
print('I will remove the content of the sim folder')
# Clean up before leave
status = my_new_mol.status()
if status == 'all_extracted': # all quantities are extracted
if debug:
print(f'status: {status}, but debug is on ==> will move {sim_folder_scratch} to {sim_folder_home}')
copytree(sim_folder_scratch, sim_folder_home, dirs_exist_ok=True) # will rewrite the folder
else:
print(f'status: {status} ==> will remove {sim_folder_scratch}')
try_to_remove_folder(sim_folder_scratch)
else:
print(f'status: {status} ==> will copy failed sim folder from scratch')
# if not os.path.exists(sim_folder_home):
# os.mkdir(sim_folder_home) # will overwrite if exists
try:
copytree(sim_folder_scratch, sim_folder_home, dirs_exist_ok=True) # will rewrite the folder > 3.8 needed
print(f"I have copied {sim_folder_scratch} to {sim_folder_home}")
except:
print(f"I could not copy {sim_folder_scratch} to {sim_folder_home}")
try_to_remove_folder(sim_folder_scratch)
def main():
scratch = os.environ['SCRATCH'] # SCRATCH has to be in the env var dict. Normally, it is.
# parser begin
parser = argparse.ArgumentParser(description='rank and num of cpus')
parser.add_argument('-rank') # array job number
parser.add_argument('-num_cpus') # number of cpus you request for every array job
parser.add_argument('-i') # input_from_yaml yaml file
args = parser.parse_args()
# parser end
# yaml file
yaml_file_name = args.i
with open(yaml_file_name) as stream:
input = yaml.load(stream=stream)
# end: yaml file
# todo: think over because it is imported twice
# end: run-or-check settings
debug = input['debug']
dummy_run = input['dummy_run']
# if not at cluster: test
# debug = True
# dummy_run = True
# end: if not at cluster
# folders names
sim = input['folder_names']['simulations']
db = input['folder_names']['database']
bh5670 = input['folder_names'][
'scratch'] # the outermost folder in the scratch folder where all other data are put
prefix_xyz_file_name = input['prefix_xyz_file_name']
my_offset = input['molecule_vacuum_offset']
try:
type_mpi = input['mpi']
except:
type_mpi = 'openmpi'
# parsing input_from_yaml
threads = int(args.num_cpus) # cpus used to compute. I do not subtract 1. This does not help
rank = '{:0>6}'.format(args.rank) # transform rank from '1' to '000001' format. This is not a general thing
xyz_file_name = f'{prefix_xyz_file_name}_{rank}.xyz'
xyz_file_location = f'{prefix_xyz_file_name}/{xyz_file_name}'
db_record_path = f'{db}/DB_{rank}.yaml' # file where the results will be saved todo: raeum es alles auf!
# check is the output exists
if os.path.exists(db_record_path):
print(f'The simulation results of mol. {rank} is already in the folder of reference')
exit()
# end: check if the output exists
if True: #not dummy_run:
sim_folder_scratch = f'{scratch}/{bh5670}/{sim}/{rank}'
else:
sim_folder_scratch = f'scratch/{bh5670}/{sim}/{rank}'
sim_folder_home = f'{sim}/{rank}' # sim folder at home exists. you create later {rank} folder
if not os.path.exists(sim_folder_scratch):
os.mkdir(sim_folder_scratch)
else:
rmtree(sim_folder_scratch) # leftovers from previous simulations will be removed
os.mkdir(sim_folder_scratch) # and the new folder will be created
# xyz object created, normal xyz file is created at scratch
try:
my_xyz_file_obj = XYZ.from_file(xyz_file_location) # object created using the file from home
except: # test
my_xyz_file_obj = XYZ.from_file('H2O.xyz') # object created using the file from home
xyz_at_scratch = sim_folder_scratch + '/' + xyz_file_name #
my_xyz_file_obj.write(xyz_at_scratch) # writes a normal xyz (into scratch)
# my molecule object is created. It will serve as a DB record
my_new_mol = Cp2kOutput(rank)
# rel_cutoff: 40; cutoff: 300; abc = 10
my_abc = str(my_xyz_file_obj.compute_box_size(offset=my_offset))[1:-2]
input['my_abc'] = my_abc
input['xyz_file_name'] = xyz_file_name
# misc
wf_corr_num_proc = 0 # 16 in the ref paper; -1 to use all
inp_file_name = 'test_2344.inp' # base file name
######################################## BEGIN: RUN CP2K TWO TIMES #####################################################
# suffix = ['2', '3', '4'] # cardinal numbers of the database
# begin: input_from_yaml
# cp2k_exe_path = '/home/artem/soft/cp2k/cp2k-7.1/exe/local/cp2k.popt'
#cp2k_exe_path = '/home/ws/bh5670/cp2k/cp2k-7.1/exe/local/cp2k.popt'
cp2k_exe_path = input['cp2k_exe_path']
my_run_type = 'mpi'
suffix = input['basis_set_suffix'] # todo: fix DZ --> 2, TZ --> 3, QZ --> 4
InputFactory.set_constants(input_from_yaml=input)
for i_bs, suffix in enumerate(suffix):
# start: ot dft
# I/O
output_file = f'out_{suffix}.out'
ot_file_name = 'OT_' + f'{suffix}_' + inp_file_name # for DFT (OT)
diag_file_name = 'DIAG_' + f'{suffix}_' + inp_file_name # for GW (DIAG)
# end: I/O
# OT dft simulation to converge quickly: create the simulation object
dft_ot_simulation = InputFactory.new_dft_ot(i_bs)
# OT dft: write input file
dft_ot_simulation.write_input_file(sim_folder_scratch + '/' + ot_file_name)
# OT dft run below ...
# ... but before, we copy the RESTART from the previous basis set (it exists unless for the smallest basis set)
if i_bs != 0:
try:
copy(sim_folder_scratch + '/' + f'{int(suffix) - 1}-RESTART.wfn',
sim_folder_scratch + '/' + f'{suffix}-RESTART.wfn')
print('copied restart file 2->3 or 3->4')
except:
print('not succesfull copy of the restart file')
elif i_bs == 0:
pass
#
print(f"Running PBE with OT (basis set = {suffix})...")
if not dummy_run:
cp2k_run(input_file=ot_file_name,
xyz_file=xyz_file_name,
run_type=my_run_type,
np=threads,
output_file=f'out_ot_{suffix}.out',
cp2k_executable=cp2k_exe_path,
execution_directory=sim_folder_scratch,
type_mpi=type_mpi)
# end: first run
print(f"I have finished cp2k with OT (basis set = {suffix})")
# DIAGONALIZATION RUN to reliably compute H**O and then GW
# gw: create the simulation object
gw_diag_simulations = InputFactory.new_gw(i_bs)
# gw: write the input file
gw_diag_simulations.write_input_file(sim_folder_scratch + '/' + diag_file_name)
# gw run
print(f"Running G0W0 with DIAG (basis set = {suffix})...")
my_out_file2 = f'out_diag_{suffix}.out'
if not dummy_run:
cp2k_run(input_file=diag_file_name,
xyz_file=xyz_file_name,
output_file=my_out_file2,
run_type=my_run_type,
np=threads,
cp2k_executable=cp2k_exe_path,
execution_directory=sim_folder_scratch,
type_mpi=type_mpi)
print(f"I have finished cp2k with DIAG (basis set = {suffix})")
# extract h**o/lumo and gw h**o/lumo from the cp2k output file:
path_to_out2_file = sim_folder_scratch + '/' + my_out_file2
# the method to extract?
# extract from the output
try:
num_orb = extract_number_of_independent_orbital_function(path_to_out2_file)
print(f'basis set = {suffix}, number of independent orbital functions: {num_orb}')
except:
print('number of orbitals was not extracted')
num_orb = 'not extracted'
try:
homos, lumos = [], []
homos, lumos = return_homo_lumo(path_to_out2_file)
print(f'basis set = {suffix} ', 'h**o = ', homos[-1]*eV_to_Hartree(), ' eV')
print(f'basis set = {suffix} ', 'lumo = ', lumos[0]*eV_to_Hartree(), ' eV')
h**o = homos[-1]*eV_to_Hartree()
lumo = lumos[0]*eV_to_Hartree()
except:
print(f'H**o/Lumo were not extracted')
h**o = 'not extracted'
lumo = 'not extracted'
try:
gw_occ, gw_vir, homo_, lumo_ = return_gw_energies(path_to_out2_file)
if isinstance(h**o, str) and isinstance(lumo, str):
h**o = homo_
lumo = lumo_
print(f'basis set = {suffix} ', 'h**o = ', h**o, ' eV')
print(f'basis set = {suffix} ', 'lumo = ', lumo, ' eV')
print(f'basis set = {suffix} ', 'gw h**o = ', gw_occ, ' eV')
print(f'basis set = {suffix} ', 'gw lumo = ', gw_vir, ' eV')
except:
print("GW energies were not extracted")
gw_occ = 'not extracted'
gw_vir = 'not extracted'
del dft_ot_simulation, gw_diag_simulations
# put computed data into the molecule object
my_new_mol.add_energies(int(suffix), h**o, lumo, gw_occ, gw_vir)
my_new_mol.add_num_orbitals(int(suffix), num_orb)
my_new_mol.extrapolate_energy() # level up?
db_record = my_new_mol.yield_dict() # this dict will be written into yaml. it will be a record in the global library
#
######################################## END: RUN CP2K TWO TIMES #######################################################
print("\nI am done\n")
if not dummy_run:
print('saving to DB...')
with open(f'{db}/DB_{rank}.yaml', 'w') as stream:
yaml.safe_dump(db_record, stream)
print(f"saved to {db}/DB_{rank}.yaml")
print('I will remove the content of the sim folder')
# Clean up before leave
status = my_new_mol.status()
if status == 'all_extracted': # all quantities are extracted
if debug:
print(f'status: {status}, but debug is on ==> will move {sim_folder_scratch} to {sim_folder_home}')
copytree(sim_folder_scratch, sim_folder_home, dirs_exist_ok=True) # will rewrite the folder
else:
print(f'status: {status} ==> will remove {sim_folder_scratch}')
try_to_remove_folder(sim_folder_scratch)
else:
print(f'status: {status} ==> will copy failed sim folder from scratch')
#if not os.path.exists(sim_folder_home):
#os.mkdir(sim_folder_home) # will overwrite if exists
try:
copytree(sim_folder_scratch, sim_folder_home) # will rewrite the folder?
print(f"I have copied {sim_folder_scratch} to {sim_folder_home}")
except:
print(f"I could not copy {sim_folder_scratch} to {sim_folder_home}")
try_to_remove_folder(sim_folder_scratch)