def cmpt_deepmd_lammps(jdata, conf_dir, task_name) :
conf_path = os.path.abspath(conf_dir)
conf_poscar = os.path.join(conf_path, 'POSCAR')
task_path = re.sub('confs', global_task_name, conf_path)
task_path = os.path.join(task_path, task_name)
equi_stress = Stress(np.loadtxt(os.path.join(task_path, 'equi.stress.out')))
lst_dfm_path = glob.glob(os.path.join(task_path, 'dfm-*'))
lst_strain = []
lst_stress = []
for ii in lst_dfm_path :
strain = np.loadtxt(os.path.join(ii, 'strain.out'))
stress = lammps.get_stress(os.path.join(ii, 'log.lammps'))
# convert from pressure to stress
stress = -stress
lst_strain.append(Strain(strain))
lst_stress.append(Stress(stress))
et = ElasticTensor.from_independent_strains(lst_strain, lst_stress, eq_stress = equi_stress, vasp = False)
# et = ElasticTensor.from_independent_strains(lst_strain, lst_stress, eq_stress = None)
# bar to GPa
# et = -et / 1e4
print_et(et)
result = os.path.join(task_path,'result')
result_et(et,conf_dir,result)
if 'upload_username' in jdata.keys() and task_name=='deepmd':
upload_username=jdata['upload_username']
util.insert_data('elastic','deepmd',upload_username,result)
def cmpt_deepmd_lammps(jdata, conf_dir, task_name):
deepmd_model_dir = jdata['deepmd_model_dir']
deepmd_type_map = jdata['deepmd_type_map']
ntypes = len(deepmd_type_map)
conf_path = os.path.abspath(conf_dir)
conf_poscar = os.path.join(conf_path, 'POSCAR')
task_path = re.sub('confs', global_task_name, conf_path)
task_path = os.path.join(task_path, task_name)
equi_stress = Stress(np.loadtxt(os.path.join(task_path,
'equi.stress.out')))
lst_dfm_path = glob.glob(os.path.join(task_path, 'dfm-*'))
lst_strain = []
lst_stress = []
for ii in lst_dfm_path:
strain = np.loadtxt(os.path.join(ii, 'strain.out'))
stress = lammps.get_stress(os.path.join(ii, 'log.lammps'))
# convert from pressure to stress
stress = -stress
lst_strain.append(Strain(strain))
lst_stress.append(Stress(stress))
et = ElasticTensor.from_independent_strains(lst_strain,
lst_stress,
eq_stress=equi_stress,
vasp=False)
# et = ElasticTensor.from_independent_strains(lst_strain, lst_stress, eq_stress = None)
# bar to GPa
# et = -et / 1e4
print_et(et)
def _compute_lower(self, output_file, all_tasks, all_res):
output_file = os.path.abspath(output_file)
res_data = {}
ptr_data = output_file + '\n'
equi_stress = Stress(
np.loadtxt(
os.path.join(os.path.dirname(output_file), 'equi.stress.out')))
lst_strain = []
lst_stress = []
for ii in all_tasks:
with open(os.path.join(ii, 'inter.json')) as fp:
idata = json.load(fp)
inter_type = idata['type']
strain = np.loadtxt(os.path.join(ii, 'strain.out'))
if inter_type == 'vasp':
stress = vasp.get_stress(os.path.join(ii, 'OUTCAR'))
# convert from pressure in kB to stress
stress *= -1000
lst_strain.append(Strain(strain))
lst_stress.append(Stress(stress))
elif inter_type in ['deepmd', 'meam', 'eam_fs', 'eam_alloy']:
stress = lammps.get_stress(os.path.join(ii, 'log.lammps'))
# convert from pressure to stress
stress = -stress
lst_strain.append(Strain(strain))
lst_stress.append(Stress(stress))
et = ElasticTensor.from_independent_strains(lst_strain,
lst_stress,
eq_stress=equi_stress,
vasp=False)
res_data['elastic_tensor'] = []
for ii in range(6):
for jj in range(6):
res_data['elastic_tensor'].append(et.voigt[ii][jj] / 1e4)
ptr_data += "%7.2f " % (et.voigt[ii][jj] / 1e4)
ptr_data += '\n'
BV = et.k_voigt / 1e4
GV = et.g_voigt / 1e4
EV = 9 * BV * GV / (3 * BV + GV)
uV = 0.5 * (3 * BV - 2 * GV) / (3 * BV + GV)
res_data['BV'] = BV
res_data['GV'] = GV
res_data['EV'] = EV
res_data['uV'] = uV
ptr_data += "# Bulk Modulus BV = %.2f GPa\n" % BV
ptr_data += "# Shear Modulus GV = %.2f GPa\n" % GV
ptr_data += "# Youngs Modulus EV = %.2f GPa\n" % EV
ptr_data += "# Poission Ratio uV = %.2f " % uV
with open(output_file, 'w') as fp:
json.dump(res_data, fp, indent=4)
return res_data, ptr_data
def cmpt_lammps(jdata, conf_dir, task_name):
strain_direct = jdata['strain_direct']
a, b = strain_direct[0], strain_direct[1]
conf_path = os.path.abspath(conf_dir)
conf_poscar = os.path.join(conf_path, 'POSCAR')
task_path = re.sub('confs', global_task_name, conf_path)
task_path = os.path.join(task_path, task_name)
lst_dfm_path = glob.glob(os.path.join(task_path, 'dfm-*'))
lst_strain = []
lst_stress = []
print('conf_dir:', conf_dir)
print('strain\t stress(kB)')
for ii in lst_dfm_path:
strain = np.loadtxt(os.path.join(ii, 'strain.out'))
stress = lammps.get_stress(os.path.join(ii, 'log.lammps'))
# convert from pressure to stress
stress = -stress / 1000
lst_strain.append(strain[a, b])
lst_stress.append(stress[a, b])
index = np.argsort(lst_strain)
for ii in range(len(lst_strain)):
print('%7.4f %7.4f' % (lst_strain[index[ii]], lst_stress[index[ii]]))
def make_lammps(jdata, conf_dir,task_type) :
fp_params = jdata['lammps_params']
model_dir = fp_params['model_dir']
type_map = fp_params['type_map']
model_dir = os.path.abspath(model_dir)
model_name =fp_params['model_name']
if not model_name and task_type =='deepmd':
models = glob.glob(os.path.join(model_dir, '*pb'))
model_name = [os.path.basename(ii) for ii in models]
assert len(model_name)>0,"No deepmd model in the model_dir"
else:
models = [os.path.join(model_dir,ii) for ii in model_name]
model_param = {'model_name' : fp_params['model_name'],
'param_type': fp_params['model_param_type']}
ntypes = len(type_map)
norm_def = jdata['norm_deform']
shear_def = jdata['shear_deform']
conf_path = os.path.abspath(conf_dir)
conf_poscar = os.path.join(conf_path, 'POSCAR')
# get equi poscar
equi_path = re.sub('confs', global_equi_name, conf_path)
equi_path = os.path.join(equi_path, task_type)
equi_dump = os.path.join(equi_path, 'dump.relax')
task_path = re.sub('confs', global_task_name, conf_path)
task_path = os.path.join(task_path, task_type)
os.makedirs(task_path, exist_ok=True)
task_poscar = os.path.join(task_path, 'POSCAR')
lammps.poscar_from_last_dump(equi_dump, task_poscar, type_map)
# get equi stress
equi_log = os.path.join(equi_path, 'log.lammps')
stress = lammps.get_stress(equi_log)
np.savetxt(os.path.join(task_path, 'equi.stress.out'), stress)
# gen strcture
# ss = Structure.from_file(conf_poscar)
# print(ss)
# ss = ss.from_file(task_poscar)
# print(ss)
ss = Structure.from_file(task_poscar)
# gen defomations
norm_strains = [-norm_def, -0.5*norm_def, 0.5*norm_def, norm_def]
shear_strains = [-shear_def, -0.5*shear_def, 0.5*shear_def, shear_def]
print('gen with norm '+str(norm_strains))
print('gen with shear '+str(shear_strains))
dfm_ss = DeformedStructureSet(ss,
symmetry = False,
norm_strains = norm_strains,
shear_strains = shear_strains)
n_dfm = len(dfm_ss)
# gen tasks
cwd = os.getcwd()
# make lammps.in
if task_type=='deepmd':
fc = lammps.make_lammps_elastic('conf.lmp',
ntypes,
lammps.inter_deepmd,
model_name)
elif task_type=='meam':
fc = lammps.make_lammps_elastic('conf.lmp',
ntypes,
lammps.inter_meam,
model_param)
f_lammps_in = os.path.join(task_path, 'lammps.in')
with open(f_lammps_in, 'w') as fp :
fp.write(fc)
cwd = os.getcwd()
os.chdir(task_path)
for ii in model_name :
if os.path.exists(ii) :
os.remove(ii)
for (ii,jj) in zip(models, model_name) :
os.symlink(os.path.relpath(ii), jj)
share_models = [os.path.join(task_path,ii) for ii in model_name]
for ii in range(n_dfm) :
# make dir
dfm_path = os.path.join(task_path, 'dfm-%03d' % ii)
os.makedirs(dfm_path, exist_ok=True)
os.chdir(dfm_path)
for jj in ['conf.lmp', 'lammps.in'] + model_name :
if os.path.isfile(jj):
os.remove(jj)
# make conf
dfm_ss.deformed_structures[ii].to('POSCAR', 'POSCAR')
lammps.cvt_lammps_conf('POSCAR', 'conf.lmp')
ptypes = vasp.get_poscar_types('POSCAR')
lammps.apply_type_map('conf.lmp', type_map, ptypes)
# record strain
strain = Strain.from_deformation(dfm_ss.deformations[ii])
np.savetxt('strain.out', strain)
# link lammps.in
os.symlink(os.path.relpath(f_lammps_in), 'lammps.in')
# link models
for (ii,jj) in zip(share_models, model_name) :
os.symlink(os.path.relpath(ii), jj)
cwd = os.getcwd()
def make_lammps(jdata, conf_dir, task_type):
fp_params = jdata['lammps_params']
model_dir = fp_params['model_dir']
type_map = fp_params['type_map']
model_dir = os.path.abspath(model_dir)
model_name = fp_params['model_name']
if not model_name:
models = glob.glob(os.path.join(model_dir, '*pb'))
model_name = [os.path.basename(ii) for ii in models]
else:
models = [os.path.join(model_dir, ii) for ii in model_name]
model_param = {
'model_name': fp_params['model_name'],
'param_type': fp_params['model_param_type']
}
ntypes = len(type_map)
strain_start = jdata['strain_start']
strain_end = jdata['strain_end']
strain_step = jdata['strain_step']
strain_direct = jdata['strain_direct']
conf_path = os.path.abspath(conf_dir)
conf_poscar = os.path.join(conf_path, 'POSCAR')
# get equi poscar
equi_path = re.sub('confs', global_equi_name, conf_path)
equi_path = os.path.join(equi_path, task_type)
equi_dump = os.path.join(equi_path, 'dump.relax')
task_path = re.sub('confs', global_task_name, conf_path)
task_path = os.path.join(task_path, task_type)
os.makedirs(task_path, exist_ok=True)
task_poscar = os.path.join(task_path, 'POSCAR')
lammps.poscar_from_last_dump(equi_dump, task_poscar, type_map)
# get equi stress
equi_log = os.path.join(equi_path, 'log.lammps')
stress = lammps.get_stress(equi_log)
np.savetxt(os.path.join(task_path, 'equi.stress.out'), stress)
# gen strcture
ss = Structure.from_file(task_poscar)
# gen defomations
norm_strains = np.arange(strain_start, strain_end, strain_step)
print('gen with norm ' + str(norm_strains))
deformations = []
for ii in norm_strains:
strain = Strain.from_index_amount(strain_direct, ii)
deformations.append(strain.get_deformation_matrix())
deformed_structures = [
defo.apply_to_structure(ss) for defo in deformations
]
n_dfm = len(deformed_structures)
# gen tasks
cwd = os.getcwd()
# make lammps.in
if task_type == 'deepmd':
fc = lammps.make_lammps_elastic('conf.lmp', ntypes,
lammps.inter_deepmd, model_name)
elif task_type == 'meam':
fc = lammps.make_lammps_elastic('conf.lmp', ntypes, lammps.inter_meam,
model_param)
f_lammps_in = os.path.join(task_path, 'lammps.in')
with open(f_lammps_in, 'w') as fp:
fp.write(fc)
cwd = os.getcwd()
if task_type == 'deepmd':
os.chdir(task_path)
for ii in model_name:
if os.path.exists(ii):
os.remove(ii)
for (ii, jj) in zip(models, model_name):
os.symlink(os.path.relpath(ii), jj)
share_models = glob.glob(os.path.join(task_path, '*pb'))
else:
share_models = models
for ii in range(n_dfm):
# make dir
dfm_path = os.path.join(task_path, 'dfm-%03d' % ii)
os.makedirs(dfm_path, exist_ok=True)
os.chdir(dfm_path)
for jj in ['conf.lmp', 'lammps.in'] + model_name:
if os.path.isfile(jj):
os.remove(jj)
# make conf
deformed_structures[ii].to('POSCAR', 'POSCAR')
lammps.cvt_lammps_conf('POSCAR', 'conf.lmp')
ptypes = vasp.get_poscar_types('POSCAR')
lammps.apply_type_map('conf.lmp', type_map, ptypes)
# record strain
strain = Strain.from_deformation(deformations[ii])
np.savetxt('strain.out', strain)
# link lammps.in
os.symlink(os.path.relpath(f_lammps_in), 'lammps.in')
# link models
for (ii, jj) in zip(share_models, model_name):
os.symlink(os.path.relpath(ii), jj)
cwd = os.getcwd()
def make_confs(self, path_to_work, path_to_equi, refine=False):
path_to_work = os.path.abspath(path_to_work)
path_to_equi = os.path.abspath(path_to_equi)
task_list = []
cwd = os.getcwd()
norm_def = self.norm_deform
shear_def = self.shear_deform
norm_strains = [-norm_def, -0.5 * norm_def, 0.5 * norm_def, norm_def]
shear_strains = [
-shear_def, -0.5 * shear_def, 0.5 * shear_def, shear_def
]
print('gen with norm ' + str(norm_strains))
print('gen with shear ' + str(shear_strains))
equi_contcar = os.path.join(path_to_equi, 'CONTCAR')
if not os.path.exists(equi_contcar):
raise RuntimeError("please do relaxation first")
ss = Structure.from_file(equi_contcar)
dfm_ss = DeformedStructureSet(ss,
symmetry=False,
norm_strains=norm_strains,
shear_strains=shear_strains)
n_dfm = len(dfm_ss)
os.chdir(path_to_work)
if os.path.isfile('POSCAR'):
os.remove('POSCAR')
os.symlink(os.path.relpath(equi_contcar), 'POSCAR')
# task_poscar = os.path.join(output, 'POSCAR')
# stress
equi_outcar = os.path.join(path_to_equi, 'OUTCAR')
equi_log = os.path.join(path_to_equi, 'log.lammps')
if os.path.exists(equi_outcar):
stress = vasp.get_stress(equi_outcar)
np.savetxt('equi.stress.out', stress)
elif os.path.exists(equi_log):
stress = lammps.get_stress(equi_log)
np.savetxt('equi.stress.out', stress)
os.chdir(cwd)
if refine:
task_list = make_refine(self.parameter['init_from_suffix'],
self.parameter['output_suffix'],
path_to_work, n_dfm)
os.chdir(cwd)
else:
for ii in range(n_dfm):
output_task = os.path.join(path_to_work, 'task.%06d' % ii)
os.makedirs(output_task, exist_ok=True)
os.chdir(output_task)
for jj in [
'INCAR', 'POTCAR', 'POSCAR', 'conf.lmp', 'in.lammps'
]:
if os.path.exists(jj):
os.remove(jj)
task_list.append(output_task)
dfm_ss.deformed_structures[ii].to('POSCAR', 'POSCAR')
# record strain
strain = Strain.from_deformation(dfm_ss.deformations[ii])
np.savetxt('strain.out', strain)
os.chdir(cwd)
return task_list