Skip to content

Commit

Permalink
Improvement of combine_imps workflow. Fix combine LDAU settings
Browse files Browse the repository at this point in the history
  • Loading branch information
@dantogni
dantogni committed Dec 9, 2023
1 parent 97a9942 commit 9cc6b32
Show file tree
Hide file tree
Showing 2 changed files with 133 additions and 44 deletions.
76 changes: 63 additions & 13 deletions aiida_kkr/tools/combine_imps.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -19,8 +19,8 @@
__copyright__ = (u'Copyright (c), 2020, Forschungszentrum Jülich GmbH, '
'IAS-1/PGI-1, Germany. All rights reserved.')
__license__ = 'MIT license, see LICENSE.txt file'
__version__ = '0.3.2'
__contributors__ = (u'Philipp Rüßmann')
__version__ = '0.3.3'
__contributors__ = (u'Philipp Rüßmann, David Antognini Silva')

# activate debug writeout
debug = False
Expand Down Expand Up @@ -262,9 +262,9 @@ def get_ldaumatrices(retrieved):
iphi = ii
ii += 1
# save matrices to output dict
txt_dict_ldaumats['wldau'] = txt[iwldau + 2:iuldau]
txt_dict_ldaumats['uldau'] = txt[iuldau + 2:iphi]
txt_dict_ldaumats['phi'] = txt[iphi + 2:]
txt_dict_ldaumats['wldau'] = txt[iwldau + 1:iuldau]
txt_dict_ldaumats['uldau'] = txt[iuldau + 1:iphi]
txt_dict_ldaumats['phi'] = txt[iphi + 1:]

return has_ldaupot_file, txt_dict_ldaumats

Expand Down Expand Up @@ -300,17 +300,11 @@ def combine_settings_ldau(**kwargs):
# now combine LDAU settings
settings_LDAU_combined = {}

if has_old_ldaupot1 or has_old_ldaupot2:
settings_LDAU_combined['initial_matrices'] = {}

if imp1_has_ldau:
for k, v in settings_LDAU1.items(): # pylint: disable=used-before-assignment
if 'iatom' in k:
iatom = int(k.split('=')[1])
# TODO: implement something for the case when LDAU is not only on the impurity site at iatom==0
settings_LDAU_combined[f'iatom={iatom}'] = v
if has_old_ldaupot1:
settings_LDAU_combined['initial_matrices'][f'iatom={iatom}'] = txts_ldaumat1 # pylint: disable=used-before-assignment

if imp2_has_ldau:
for k, v in settings_LDAU2.items(): # pylint: disable=used-before-assignment
Expand All @@ -321,7 +315,63 @@ def combine_settings_ldau(**kwargs):
else:
noffset = kickout_info['Ncls1']
settings_LDAU_combined[f'iatom={iatom+noffset}'] = v
if has_old_ldaupot2:
settings_LDAU_combined['initial_matrices'][f'iatom={iatom+noffset}'] = txts_ldaumat2 # pylint: disable=used-before-assignment

if has_old_ldaupot1:
settings_LDAU_combined['initial_matrices'] = {}
settings_LDAU_combined['initial_matrices'] = get_LDAU_initmatrices_dict(txts_ldaumat1) # pylint: disable=used-before-assignment

if has_old_ldaupot2:
# If the dictionary already exists, update it with the new values
if 'initial_matrices' in settings_LDAU_combined:
settings_LDAU_combined['initial_matrices'].update(get_LDAU_initmatrices_dict(txts_ldaumat2, noffset))
else:
# If the dictionary does not exist, create it with the new values
settings_LDAU_combined['initial_matrices'] = {}
settings_LDAU_combined['initial_matrices'] = get_LDAU_initmatrices_dict(txts_ldaumat2, noffset)

return Dict(settings_LDAU_combined)


def get_LDAU_initmatrices_dict(txts_ldaumat1, offset=0):
"""
Put the initial matrices text content of `get_ldaumatrices` output in a
Dictionnary form sorted by atoms.
"""

# Loop through characters ('wldau', 'uldau', 'phi')
for char in ['wldau', 'uldau', 'phi']:
atom_line = []
iatoms = []

# Save atom number and lines in which atoms information starts
for iline in range(len(txts_ldaumat1[char])):
if txts_ldaumat1[char][iline][:4] == 'atom':
iatom = txts_ldaumat1[char][iline].split()[1]
iatoms.append(iatom)
atom_line.append(iline)
atom_line.append(len(txts_ldaumat1[char]))

mat = []
# Extract matrices based on atom lines
for iatom in range(len(atom_line) - 1):
mat.append(txts_ldaumat1[char][atom_line[iatom] + 1:atom_line[iatom + 1]])

# Assign matrices to corresponding variables based on the character
if char == 'wldau':
wldaumat = mat
elif char == 'uldau':
uldaumat = mat
elif char == 'phi':
phimat = mat

# Create a dictionary to store the organized data
LDAU_initmatrices_dict = {}

# Fill the dictionary with atom-wise information
for iatom in range(len(iatoms)):
LDAU_initmatrices_dict[f'iatom={int(iatoms[iatom])-1+offset}'] = {}
LDAU_initmatrices_dict[f'iatom={int(iatoms[iatom])-1+offset}']['wldau'] = wldaumat[iatom]
LDAU_initmatrices_dict[f'iatom={int(iatoms[iatom])-1+offset}']['uldau'] = uldaumat[iatom]
LDAU_initmatrices_dict[f'iatom={int(iatoms[iatom])-1+offset}']['phi'] = phimat[iatom]

return LDAU_initmatrices_dict
101 changes: 70 additions & 31 deletions aiida_kkr/workflows/_combine_imps.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -19,11 +19,11 @@
__copyright__ = (u'Copyright (c), 2020, Forschungszentrum Jülich GmbH, '
'IAS-1/PGI-1, Germany. All rights reserved.')
__license__ = 'MIT license, see LICENSE.txt file'
__version__ = '0.3.2'
__contributors__ = (u'Philipp Rüßmann , Rubel Mozumder')
__version__ = '0.3.3'
__contributors__ = (u'Philipp Rüßmann , Rubel Mozumder, David Antognini Silva')

# activate debug writeout
_debug = False
_debug = True


class combine_imps_wc(WorkChain):
Expand Down Expand Up @@ -106,7 +106,9 @@ def define(cls, spec):

# expose these inputs from sub-workflows
spec.expose_inputs(
kkr_imp_sub_wc, namespace='scf', include=('kkrimp', 'options', 'wf_parameters', 'params_overwrite')
kkr_imp_sub_wc,
namespace='scf',
include=('kkrimp', 'options', 'wf_parameters', 'settings_LDAU', 'params_overwrite')
)
spec.expose_inputs(
kkr_flex_wc,
Expand Down Expand Up @@ -276,28 +278,30 @@ def combine_single_single(self):
imp_1 = self.ctx.imp1
imp_2 = self.ctx.imp2
# check for the impurity1 whether from single kkr_imp_wc or not
if imp_1.process_class == KkrCalculation:
if imp_1.process_class == KkrCalculation or imp_1.process_class == KkrimpCalculation:
Zimp_num_1 = imp_1.inputs.impurity_info.get_dict().get('Zimp')
if isinstance(Zimp_num_1, list):
if len(Zimp_num_1) > 1:
single_imp_1 = False

elif imp_1.process_class == kkr_imp_sub_wc:
combine_wc = imp_1.get_incoming(node_class=combine_imps_wc).all()
if len(combine_wc) != 0:
Zimp_num_1 = imp_1.inputs.impurity_info.get_dict().get('Zimp')
if len(Zimp_num_1) > 1:
single_imp_1 = False
elif imp_1.process_class == combine_imps_wc:
single_imp_1 = False

# check for the impurity2 whether from single kkr_imp_wc or not
if imp_2.process_class == KkrCalculation:
if imp_2.process_class == KkrCalculation or imp_2.process_class == KkrimpCalculation:
Zimp_num_2 = imp_2.inputs.impurity_info.get_dict().get('Zimp')
if isinstance(Zimp_num_2, list):
if len(Zimp_num_2) > 1:
single_imp_2 = False
elif imp_2.process_class == kkr_imp_sub_wc:
combine_wc = imp_2.get_incoming(node_class=combine_imps_wc).all()
if len(combine_wc) != 0:
Zimp_num_2 = imp_2.inputs.impurity_info.get_dict().get('Zimp')
if len(Zimp_num_2) > 1:
single_imp_2 = False
elif imp_2.process_class == combine_imps_wc:
single_imp_2 = False
Expand Down Expand Up @@ -337,18 +341,19 @@ def extract_imps_info_exact_cluster(self):
# This 'if clause' to extract the imps_info_in_exact_cluster from workflow_info of the input impurity node
if imp1_input.process_class == combine_imps_wc:
parent_combine_wc = imp1_input
out_workflow_info = parent_combine_wc.outputs.workflow_info


# out_workflow_info = parent_combine_wc.get_outgoing(link_label_filter='workflow_info').all()[0].node

elif imp1_input.process_class == KkrimpCalculation:
kkrimp_sub = imp1_input.get_incoming(node_class=kkr_imp_sub_wc).all()[0].node
parent_combine_wc = kkrimp_sub.get_incoming(node_class=combine_imps_wc).all()[0].node
out_workflow_info = imp1_input.outputs.output_parameters

elif imp1_input.process_class == kkr_imp_sub_wc:
parent_combine_wc = imp1_input.get_incoming(node_class=combine_imps_wc).all()[0].node

out_workflow_info = parent_combine_wc.outputs.workflow_info
out_workflow_info = imp1_input.outputs.workflow_info

imp2_impurity_info = self.ctx.imp2.inputs.impurity_info

Expand Down Expand Up @@ -385,7 +390,10 @@ def get_impinfo_from_hostGF(self, imp_calc):
"""
Extract impurity infor node from the incoming host GF folder
"""
GF_input = imp_calc.inputs.host_Greenfunction_folder
if imp_calc.process_class == KkrimpCalculation:
GF_input = imp_calc.inputs.host_Greenfunction_folder
elif imp_calc.process_class == kkr_imp_sub_wc:
GF_input = imp_calc.inputs.remote_data
parent_calc = GF_input.get_incoming(node_class=CalcJobNode).first().node
impinfo = parent_calc.inputs.impurity_info
return impinfo
Expand Down Expand Up @@ -494,13 +502,18 @@ def create_big_cluster(self): # pylint: disable=inconsistent-return-statements
if _debug:
print('DEBUG:', list(imp1.inputs))
impinfo1 = self.get_impinfo_from_hostGF(imp1)
impinfo2 = self.get_impinfo_from_hostGF(imp2)
# impinfo1 = imp1.inputs.impurity_info
else:
if imp1.process_class == self.__class__:
imp1 = imp1.get_outgoing(node_class=kkr_imp_sub_wc).all()[0].node
impinfo1 = imp1.inputs.impurity_info
self.report(f'DEBUG: impinfo1 : {impinfo1.get_dict()} .')
impinfo2 = self.get_impinfo_from_hostGF(imp2)

if imp2.process_class == self.__class__:
imp2 = imp2.get_outgoing(node_class=kkr_imp_sub_wc).all()[0].node
impinfo2 = imp2.inputs.impurity_info
self.report(f'DEBUG: impinfo2 : {impinfo2.get_dict()} .')

host_structure = self.ctx.host_structure
offset_imp2 = self.inputs.offset_imp2
Expand All @@ -515,8 +528,9 @@ def create_big_cluster(self): # pylint: disable=inconsistent-return-statements
if single_single:
if offset_imp2.get_dict()['index'] < 0:
return self.exit_codes.ERROR_INPLANE_NEIGHBOR_TOO_SMALL # pylint: disable=maybe-no-member
if impinfo1['ilayer_center'] == impinfo2['ilayer_center'] and self.inputs.offset_imp2['index'] < 1:
return self.exit_codes.ERROR_INPLANE_NEIGHBOR_TOO_SMALL # pylint: disable=maybe-no-member
if 'ilayer_center' in impinfo1.get_dict() and 'ilayer_center' in impinfo2.get_dict():
if impinfo1['ilayer_center'] == impinfo2['ilayer_center'] and self.inputs.offset_imp2['index'] < 1:
return self.exit_codes.ERROR_INPLANE_NEIGHBOR_TOO_SMALL # pylint: disable=maybe-no-member
else:
imp_offset_index = offset_imp2['index']
imp2_ilayer = impinfo2['ilayer_center']
Expand Down Expand Up @@ -755,6 +769,9 @@ def run_kkrimp_scf(self):
if add_ldausettings:
self.report(f'settings_combined: {settings_LDAU_combined.get_dict()}')
builder.settings_LDAU = settings_LDAU_combined
if 'settings_LDAU' in self.inputs.scf:
self.report('Overwriting settings_LDAU_combined with provided settings_LDAU inputs')
builder.settings_LDAU = self.inputs.scf.settings_LDAU

# now submit workflow
future = self.submit(builder)
Expand Down Expand Up @@ -806,38 +823,60 @@ def get_ldau_combined(self):
check if impurity input calculations have LDA+U settings in input and add this here if needed
"""

imp1_has_ldau = 'settings_LDAU' in self.ctx.imp1.inputs
if imp1_has_ldau:
settings_LDAU1 = self.ctx.imp1.inputs.settings_LDAU
self.report('found LDA+U settings for imp1')

imp2_has_ldau = 'settings_LDAU' in self.ctx.imp2.inputs
if imp2_has_ldau:
settings_LDAU2 = self.ctx.imp2.inputs.settings_LDAU
self.report('found LDA+U settings for imp1')
if self.ctx.imp1.process_class == KkrimpCalculation or self.ctx.imp1.process_class == kkr_imp_sub_wc:
imp1_has_ldau = 'settings_LDAU' in self.ctx.imp1.inputs
if imp1_has_ldau:
settings_LDAU1 = self.ctx.imp1.inputs.settings_LDAU
self.report('found LDA+U settings for imp1')
if self.ctx.imp1.process_class == KkrimpCalculation:
retrieved1 = self.ctx.imp1.outputs.retrieved
elif self.ctx.imp1.process_class == kkr_imp_sub_wc:
retrieved1 = self.ctx.imp1.get_outgoing(node_class=KkrimpCalculation).first().node.outputs.retrieved
elif self.ctx.imp1.process_class == combine_imps_wc:
imp1_has_ldau = 'settings_LDAU' in self.ctx.imp1.get_outgoing(node_class=kkr_imp_sub_wc).first().node.inputs
if imp1_has_ldau:
settings_LDAU1 = self.ctx.imp1.get_outgoing(node_class=kkr_imp_sub_wc).first().node.inputs.settings_LDAU
self.report('found LDA+U settings for imp1')
retrieved1 = self.ctx.imp1.get_outgoing(node_class=kkr_imp_sub_wc
).first().node.get_outgoing(node_class=KkrimpCalculation
).first().node.outputs.retrieved

if self.ctx.imp2.process_class == KkrimpCalculation or self.ctx.imp2.process_class == kkr_imp_sub_wc:
imp2_has_ldau = 'settings_LDAU' in self.ctx.imp2.inputs
if imp2_has_ldau:
settings_LDAU2 = self.ctx.imp2.inputs.settings_LDAU
self.report('found LDA+U settings for imp2')
if self.ctx.imp2.process_class == KkrimpCalculation:
retrieved2 = self.ctx.imp2.outputs.retrieved
elif self.ctx.imp2.process_class == kkr_imp_sub_wc:
retrieved2 = self.ctx.imp2.get_outgoing(node_class=KkrimpCalculation).first().node.outputs.retrieved
elif self.ctx.imp2.process_class == combine_imps_wc:
imp2_has_ldau = 'settings_LDAU' in self.ctx.imp2.get_outgoing(node_class=kkr_imp_sub_wc).first().node.inputs
if imp2_has_ldau:
settings_LDAU2 = self.ctx.imp2.get_outgoing(node_class=kkr_imp_sub_wc).first().node.inputs.settings_LDAU
self.report('found LDA+U settings for imp2')
retrieved2 = self.ctx.imp2.get_outgoing(node_class=kkr_imp_sub_wc
).first().node.get_outgoing(node_class=KkrimpCalculation
).first().node.outputs.retrieved

if imp1_has_ldau and imp2_has_ldau:
# combine LDA+U settings of the two imps
settings_LDAU_combined = combine_settings_ldau(
settings_LDAU1=settings_LDAU1,
retrieved1=self.ctx.imp1.outputs.retrieved,
retrieved1=retrieved1,
settings_LDAU2=settings_LDAU2,
retrieved2=self.ctx.imp2.outputs.retrieved,
retrieved2=retrieved2,
kickout_info=self.ctx.kickout_info
)
elif imp1_has_ldau:
# use only LDA+U settings of imp 1
settings_LDAU_combined = combine_settings_ldau(
settings_LDAU1=settings_LDAU1,
retrieved1=self.ctx.imp1.outputs.retrieved,
kickout_info=self.ctx.kickout_info
settings_LDAU1=settings_LDAU1, retrieved1=retrieved1, kickout_info=self.ctx.kickout_info
)
elif imp2_has_ldau:
# add offset to atom index for combined LDA+U settings
settings_LDAU_combined = combine_settings_ldau(
settings_LDAU2=settings_LDAU2,
retrieved2=self.ctx.imp2.outputs.retrieved,
kickout_info=self.ctx.kickout_info
settings_LDAU2=settings_LDAU2, retrieved2=retrieved2, kickout_info=self.ctx.kickout_info
)
else:
# return builder unchanged if none of the impurt calculations has LDA+U settings
Expand Down

0 comments on commit 9cc6b32

Please sign in to comment.