def h2_pyscf_equil_test():
jobs = []
xyz = open('structures/h2_equil.xyz', 'r')
xyz.readline() # Remove headers.
xyz.readline()
scfman = PySCFManager(name='scf',
path='test_h2eq_py',
writer=PySCFWriter({
'spin': 0,
'xyz': xyz.read()
}),
runner=PySCFRunnerPBS(queue='secondary',
nn=1,
np=16,
walltime='0:01:00'))
jobs.append(scfman)
var = QWalkManager(name='var',
path=scfman.path,
writer=VarianceWriter(),
reader=VarianceReader(),
runner=RunnerPBS(nn=1,
np=16,
queue='secondary',
walltime='0:05:00'),
trialfunc=SlaterJastrow(scfman, kpoint=0))
jobs.append(var)
lin = QWalkManager(name='linear',
path=scfman.path,
writer=LinearWriter(),
reader=LinearReader(),
runner=RunnerPBS(nn=1,
np=16,
queue='secondary',
walltime='0:10:00'),
trialfunc=SlaterJastrow(slatman=scfman,
jastman=var,
kpoint=0))
jobs.append(lin)
dmc = QWalkManager(name='dmc',
path=scfman.path,
writer=DMCWriter(),
reader=DMCReader(),
runner=RunnerPBS(
nn=1,
np=16,
queue='secondary',
walltime='0:10:00',
),
trialfunc=SlaterJastrow(slatman=scfman, jastman=lin))
jobs.append(dmc)
return jobs
def si_pyscf_test():
''' Simple tests that check PBC is working Crystal, and that QMC can be performed on the result.'''
jobs=[]
pwriter=PySCFPBCWriter({
'gmesh':[16,16,16],
'cif':open('si.cif','r').read()
})
pman=PySCFManager(
name='scf',
path='sipyscf',
writer=pwriter,
runner=PySCFRunnerPBS(
queue='secondary',
nn=1,
np=16,
walltime='4:00:00'
)
)
jobs.append(pman)
var=QWalkManager(
name='var',
path=pman.path,
writer=VarianceWriter(),
reader=VarianceReader(),
runner=RunnerPBS(
nn=1,queue='secondary',walltime='0:10:00'
),
trialfunc=SlaterJastrow(pman,kpoint=0)
)
jobs.append(var)
lin=QWalkManager(
name='lin',
path=pman.path,
writer=LinearWriter(),
reader=LinearReader(),
runner=RunnerPBS(
nn=1,queue='secondary',walltime='0:30:00'
),
trialfunc=SlaterJastrow(slatman=pman,jastman=var,kpoint=0)
)
jobs.append(lin)
return jobs
def si_crystal_test():
''' Simple tests that check PBC is working Crystal, and that QMC can be performed on the result.'''
jobs=[]
cwriter=CrystalWriter({
'xml_name':'../BFD_Library.xml',
'cutoff':0.2,
'kmesh':(4,4,4),
})
cwriter.set_struct_fromcif(open('si.cif','r').read(),primitive=True)
cwriter.set_options({})
cman=CrystalManager(
name='crys',
path='sicrys',
writer=cwriter,
runner=RunnerPBS(
queue='secondary',
nn=1,
walltime='0:10:00'
)
)
jobs.append(cman)
var=QWalkManager(
name='var',
path=cman.path,
writer=VarianceWriter(),
reader=VarianceReader(),
runner=RunnerPBS(
nn=1,queue='secondary',walltime='0:10:00'
),
trialfunc=SlaterJastrow(cman,kpoint=0)
)
jobs.append(var)
lin=QWalkManager(
name='linear',
path=cman.path,
writer=LinearWriter(),
reader=LinearReader(),
runner=RunnerPBS(
nn=1,queue='secondary',walltime='1:00:00'
),
trialfunc=SlaterJastrow(slatman=cman,jastman=var,kpoint=0)
)
jobs.append(lin)
for kidx in cman.qwfiles['kpoints']:
kpt=cman.qwfiles['kpoints'][kidx]
dmc=QWalkManager(
name='dmc_%d'%kidx,
path=cman.path,
writer=DMCWriter(),
reader=DMCReader(),
runner=RunnerPBS(
nn=1,queue='secondary',walltime='1:00:00',
),
trialfunc=SlaterJastrow(slatman=cman,jastman=lin,kpoint=kidx)
)
jobs.append(dmc)
return jobs
def mno_test():
''' Test spinful calculations in PBC and real kpoint sampling. Also test trialfunc dependency. '''
jobs=[]
cwriter=CrystalWriter({
'xml_name':'../BFD_Library.xml',
'cutoff':0.2,
'kmesh':(4,4,4),
'total_spin':4
})
cwriter.set_struct_fromcif(open('mno.cif','r').read(),primitive=True)
cwriter.set_options({})
cman=CrystalManager(
name='crys',
path='mnocrys',
writer=cwriter,
runner=RunnerPBS(
queue='secondary',
nn=1,
walltime='1:00:00'
)
)
jobs+=[cman]
var=QWalkManager(
name='var',
path=cman.path,
writer=VarianceWriter(),
reader=VarianceReader(),
runner=RunnerPBS(
nn=1,queue='secondary',walltime='0:10:00'
),
trialfunc=SlaterJastrow(cman,kpoint=(0,0,0))
)
lin=QWalkManager(
name='linear',
path=cman.path,
writer=LinearWriter(),
reader=LinearReader(),
runner=RunnerPBS(
nn=1,queue='secondary',walltime='1:00:00'
),
trialfunc=SlaterJastrow(slatman=cman,jastman=var,kpoint=(0,0,0))
)
for kidx in cman.qwfiles['kpoints']:
kpt=cman.qwfiles['kpoints'][kidx]
if not sum(np.array(kpt)%(np.array(cman.kmesh)/2))<1e-14:
print("complex kpt %d%d%d skipped"%kpt)
continue
dmc=QWalkManager(
name='dmc_%d'%kidx,
path=cman.path,
writer=DMCWriter(),
reader=DMCReader(),
runner=RunnerPBS(
nn=1,queue='secondary',walltime='1:00:00',
),
trialfunc=SlaterJastrow(slatman=cman,jastman=lin,kpoint=kidx)
)
jobs.append(dmc)
return jobs
def mno_crystal_test():
''' Test spinful calculations in PBC and real kpoint sampling. Also test trialfunc dependency. '''
cwriter = CrystalWriter({
'xml_name': '../../BFD_Library.xml',
'cutoff': 0.2,
'kmesh': (3, 3, 3),
'initial_spins': [1, 0],
'total_spin': 5
})
cwriter.set_struct_fromcif(open('structures/mno.cif', 'r').read(),
primitive=True)
cwriter.set_options({})
cman = CrystalManager(name='crys',
path='test_mno_crys',
writer=cwriter,
runner=RunnerPBS(queue='secondary',
nn=1,
np=20,
walltime='2:00:00'))
var = QWalkManager(name='var',
path=cman.path,
writer=VarianceWriter(),
reader=VarianceReader(),
runner=RunnerPBS(nn=1,
np=20,
queue='secondary',
walltime='0:10:00'),
trialfunc=SlaterJastrow(cman, kpoint=0))
lin = QWalkManager(name='linear',
path=cman.path,
writer=LinearWriter(),
reader=LinearReader(),
runner=RunnerPBS(nn=1,
np=20,
queue='secondary',
walltime='1:00:00'),
trialfunc=SlaterJastrow(slatman=cman,
jastman=var,
kpoint=0))
jobs = [cman, var, lin]
for kidx in cman.qwfiles['kpoints']:
kpt = cman.qwfiles['kpoints'][kidx]
dmc = QWalkManager(name='dmc_%d' % kidx,
path=cman.path,
writer=DMCWriter(),
reader=DMCReader(),
runner=RunnerPBS(
nn=1,
np=16,
queue='secondary',
walltime='1:00:00',
),
trialfunc=SlaterJastrow(slatman=cman,
jastman=lin,
kpoint=kidx))
jobs.append(dmc)
return jobs
def h2_crystal_equil_test():
jobs = []
cwriter = CrystalWriter({
'xml_name': '../../BFD_Library.xml',
'cutoff': 0.2,
'spin_polarized': False
})
cwriter.set_struct_fromxyz(open('structures/h2_equil.xyz', 'r').read())
cman = CrystalManager(name='crys',
path='test_h2eq_crys',
writer=cwriter,
runner=RunnerPBS(queue='secondary',
nn=1,
np=16,
walltime='0:01:00'))
jobs.append(cman)
var = QWalkManager(name='var',
path=cman.path,
writer=VarianceWriter(),
reader=VarianceReader(),
runner=RunnerPBS(nn=1,
np=16,
queue='secondary',
walltime='0:05:00'),
trialfunc=SlaterJastrow(cman, kpoint=0))
jobs.append(var)
lin = QWalkManager(name='linear',
path=cman.path,
writer=LinearWriter(),
reader=LinearReader(),
runner=RunnerPBS(nn=1,
np=16,
queue='secondary',
walltime='0:10:00'),
trialfunc=SlaterJastrow(slatman=cman,
jastman=var,
kpoint=0))
jobs.append(lin)
for kidx in cman.qwfiles['kpoints']:
kpt = cman.qwfiles['kpoints'][kidx]
dmc = QWalkManager(name='dmc_%d' % kidx,
path=cman.path,
writer=DMCWriter(),
reader=DMCReader(),
runner=RunnerPBS(
nn=1,
np=16,
queue='secondary',
walltime='0:10:00',
),
trialfunc=SlaterJastrow(slatman=cman,
jastman=lin,
kpoint=kidx))
jobs.append(dmc)
return jobs
writer=PySCFWriter({
'xyz': h2,
'method': 'RKS',
'dft': 'pbe,pbe'
}),
runner=PySCFRunnerLocal())
# The PySCF calculation is defined, but so far no calculations have been started.
#
# QWalk writers can take explicit trial wave function input (as a string), but autogen can also handle it for you.
# Make a `TrialFunc` object with the manager you want to generate the trial wave function.
# In[ ]:
from trialfunc import SlaterJastrow
sj = SlaterJastrow(pyscf_manager)
# ### Setting up the QWalk run.
# Now we can set up the corresponding objects for QWalk.
# In[ ]:
from variance import VarianceWriter, VarianceReader
from autorunner import RunnerLocal
from qwalkmanager import QWalkManager
from imp import reload
import time
vopt_writer = VarianceWriter({'iterations': 6})
vopt_manager = QWalkManager(
})
vmc_writer = VarianceWriter()
# Manage the PySCF job.
scfmanager = PySCFManager(name='h2_pbe',
path=path,
writer=pyscf_writer,
runner=PySCFRunnerLocal())
# Manage the VMC job.
voptmanager = QWalkManager(name='h2_vopt',
path=path,
writer=VarianceWriter(),
reader=VarianceReader(),
trialfunc=SlaterJastrow(scfmanager),
runner=RunnerLocal(np=4))
# In[ ]:
while not voptmanager.completed:
voptmanager.nextstep()
# In[ ]:
import matplotlib.pyplot as plt
get_ipython().run_line_magic('matplotlib', 'inline')
fig, ax = plt.subplots(1, 1)
ax.plot(voptmanager.reader.output['sigma'])
ax.set_title('VMC variance optimization')
ax.set_xlabel('VMC step')