def getOptims():
linopt1 = linear(
energy=0.95,
unreweightedvariance=0.0,
reweightedvariance=0.05,
timestep=0.3,
samples=12800,
walkers=1,
warmupsteps=10,
blocks=100,
steps=1,
substeps=3,
#gpu = True,
# Disable nonlocalpp for GPU, i.e., nonlocalpp = False,
nonlocalpp=True,
maxweight=1e9,
minmethod='OneShiftOnly',
minwalkers=0.4,
usebuffer=True,
exp0=-2,
bigchange=10.0,
alloweddifference=6e-04,
stepsize=0.05,
nstabilizers=3,
)
optims = []
optims.append(loop(max=5, qmc=linopt1))
return optims
def gamma_opt_input(p2q, system, init_jas=None):
from nexus import loop, linear
myid = p2q.identifier.replace('-p2q', '-opt')
nscf_dir = os.path.basename(p2q.path)
mypath = p2q.path.replace(nscf_dir, 'opt')
linopt = obj(
minmethod='OneShiftOnly',
energy=0.95,
reweightedvariance=0.05,
unreweightedvariance=0.0,
warmupsteps=40,
blocks=128,
substeps=3,
timestep=0.8,
samples=8192,
)
calcs = [loop(max=5, qmc=linear(**linopt))]
myjas = init_jas
if init_jas is None:
myjas = [('J1', 'size', 8, 'cusp', 1), ('J2', 'size', 8)]
# end if
opt_inputs = obj(
identifier=myid,
path=mypath,
input_type='basic',
system=system,
bconds='ppp', # periodic in xyz directions
calculations=calcs,
twistnum=0,
estimators=[],
jastrows=myjas,
pseudos=[],
dependencies=[(p2q, 'orbitals')])
return opt_inputs
det_format = 'old',
identifier = 'opt',
path = 'opt',
job = job(cores=16,threads=16,app=qmcapp),
input_type = 'basic',
spin_polarized = True,
system = dia,
pseudos = ['C.BFD.xml'],
jastrows = [('J1','bspline',8),('J2','bspline',8)],
calculations = [
loop(max=4,
qmc = linear(
minmethod = 'OneShiftOnly',
minwalkers = 0.5,
samples = 5000,
walkers = 1,
warmupsteps = 20,
blocks = 25,
substeps = 5,
timestep = .4,
),
),
],
dependencies = (conv,'orbitals'),
)
qmc_ground = generate_qmcpack(
det_format = 'old',
identifier = 'vmc',
path = 'vmc_ground',
job = job(cores=16,threads=16,app=qmcapp),
input_type = 'basic',
system = graphene, # run graphene
# input format selector
input_type = 'basic',
# qmcpack input parameters
corrections = [],
jastrows = [('J1','bspline',8), # 1 body bspline jastrow
('J2','bspline',8)], # 2 body bspline jastrow
calculations = [
loop(max = 6, # No. of loop iterations
qmc = linear( # linearized optimization method
energy = 0.0, # cost function
unreweightedvariance = 1.0, # is all unreweighted variance
reweightedvariance = 0.0, # no energy or r.w. var.
timestep = 0.5, # vmc timestep (1/Ha)
warmupsteps = 100, # MC steps before data collected
samples = 16000,# samples used for cost function
stepsbetweensamples = 10, # steps between uncorr. samples
blocks = 10, # ignore this
minwalkers = 0.1,# and this
bigchange = 15.0,# and this
alloweddifference = 1e-4 # and this, for now
)
)
],
# workflow dependencies
dependencies = (p2q_opt,'orbitals'),
)
# nscf run to produce orbitals for final dmc
nscf = generate_pwscf(
# nexus inputs
('J2', 'bspline', 8, 10.0)
],
calculations=[
loop(max=12,
qmc=linear(
energy=0.0,
unreweightedvariance=1.0,
reweightedvariance=0.0,
timestep=0.3,
samples=61440,
warmupsteps=50,
blocks=200,
substeps=1,
nonlocalpp=True,
usebuffer=True,
walkers=1,
minwalkers=0.5,
maxweight=1e9,
usedrift=False,
minmethod='quartic',
beta=0.025,
exp0=-16,
bigchange=15.0,
alloweddifference=1e-4,
stepsize=0.2,
stabilizerscale=1.0,
nstabilizers=3,
))
],
dependencies=(p2q, 'orbitals'),
)
def mworkflow(shared_qe, params, sims):
for i in params_defaults.keys():
if i not in params:
setattr(params, i, params_defaults[i])
#end if
#end if
if params.j3:
if params.code == 'gpu':
print "GPU code and j3 not possible yet"
exit()
#end if
#end if
if params.excitation is not None:
params.twistnum = 0
########################
#DMC and VMC parameters#
########################
samples = 25600 #25600
vmcblocks = 100
dmcblocks = 300
dmceqblocks = 200
dmcwalkers = 1024 # For cades it is multiplied by 2
vmcdt = 0.3
if params.vmc:
vmcblocks = 9000
vmcdt = 0.3
dmcblocks = 1
dmceqblocks = 1
#end if
#############################
# Machine specific variables#
#############################
if params.machine == 'cades':
minnodes = 1
scf_nodes = 2
scf_hours = 48
p2q_nodes = scf_nodes
p2q_hours = 1
qeapp = 'pw.x -npool 4 '
if params.dft_grid == (1, 1, 1):
scf_nodes = 1
qeapp = 'pw.x'
#end if
qe_presub = 'module purge; module load PE-intel/3.0; module load hdf5_parallel/1.10.3'
opt_nodes = 4
opt_hours = 12
dmcnodespertwist = 0.5
dmc_hours = 24
qmc_threads = 18
walkers = qmc_threads
blocks = samples / (walkers * opt_nodes)
params.code = 'cpu'
qmcapp = ' qmcpack_cades_cpu_comp_SoA'
qmc_presub = 'module purge; module load PE-intel'
opt_job = Job(nodes=opt_nodes,
hours=opt_hours,
threads=qmc_threads,
app=qmcapp,
presub=qmc_presub)
elif params.machine == 'titan' or params.machine == 'eos':
minnodes = 1
dmcwalkers *= 2
scf_nodes = 8
scf_hours = 2
p2q_nodes = scf_nodes
p2q_hours = 1
qeapp = 'pw.x -npool 4 '
qe_presub = ''
if params.machine == 'eos':
opt_nodes = 16
opt_hours = 2
qmc_threads = 16
walkers = qmc_threads
blocks = samples / (walkers * opt_nodes)
params.code = 'cpu'
qmcapp = '/ccs/home/kayahan/SOFTWARE/qmcpack/eos/qmcpack/qmcpack_eos_cpu_comp_SoA'
qmc_presub = ''
params.rundmc = False
elif params.machine == 'titan':
if params.code == 'cpu':
dmcwalkers *= 2
vmcblocks = 50
opt_nodes = 16
opt_hours = 2
dmcnodespertwist = 32
dmc_hours = 6
qmc_threads = 16
walkers = qmc_threads
blocks = samples / (walkers * opt_nodes)
qmcapp = '/ccs/home/kayahan/SOFTWARE/qmcpack/titan/qmcpack/qmcpack_titan_cpu_comp_SoA'
qmc_presub = ''
params.rundmc = True
elif params.code == 'gpu':
dmcwalkers *= 8
opt_nodes = 4
opt_hours = 4
dmcnodespertwist = 8
dmc_hours = 1.5
qmc_threads = 16
walkers = qmc_threads
blocks = samples / (walkers * opt_nodes)
qmcapp = '/ccs/home/kayahan/SOFTWARE/qmcpack/titan/qmcpack/qmcpack_titan_gpu_comp_SoA'
qmc_presub = 'module load cudatoolkit'
#end if
#end if
opt_job = Job(nodes=opt_nodes,
hours=opt_hours,
threads=qmc_threads,
app=qmcapp,
presub=qmc_presub)
elif params.machine == 'cetus' or params.machine == 'mira':
if params.machine == 'cetus':
minnodes = 128
opt_hours = 1
dmc_hours = 1
elif params.machine == 'mira':
minnodes = 128
dmc_hours = 1
dmc_hours = 1
#end if
scf_nodes = 128
scf_hours = 1
p2q_nodes = 128
p2q_hours = 1
qeapp = 'pw.x'
qe_presub = ''
opt_nodes = 128.0
opt_nodes = np.round(opt_nodes / minnodes) * minnodes
dmcnodespertwist = 18
qmc_threads = 16
walkers = qmc_threads
qmc_processes_per_node = 16
blocks = samples / (walkers * opt_nodes)
params.code = 'cpu'
qmcapp = 'qmcpack'
qmc_presub = ''
opt_job = Job(nodes=opt_nodes,
hours=opt_hours,
threads=qmc_threads,
processes_per_node=qmc_processes_per_node,
app=qmcapp,
presub=qmc_presub)
elif params.machine == 'summit':
dmcwalkers *= 8
minnodes = 1
scf_nodes = 2
scf_hours = 24
p2q_nodes = 2
p2q_hours = 1
qeapp = 'pw.x -npool 4 '
if params.dft_grid == (1, 1, 1):
scf_nodes = 1
qeapp = 'pw.x'
#end if
qe_presub = 'module purge; module load PE-intel/3.0'
opt_nodes = 4
opt_hours = 4
dmcnodespertwist = 8
dmc_hours = 6
qmc_threads = 42
walkers = qmc_threads
blocks = samples / (walkers * opt_nodes)
qmcapp = '/ccs/home/kayahan/SOFTWARE/qmcpack/summit/qmcpack/build_summit_comp/bin/qmcpack'
else:
print "Machine is not defined!"
exit()
#end if
scf_job = Job(nodes=scf_nodes,
hours=scf_hours,
app=qeapp,
presub=qe_presub)
p2q_job = Job(nodes=p2q_nodes,
hours=p2q_hours,
app='pw2qmcpack.x -npool 4 ',
presub=qe_presub)
if params.big or params.qp:
dmcblocks *= 4
dmcnodespertwist *= 1
dmc_hours *= 2
dmcwalkers *= 2
#end if
shared_qe.wf_collect = False
shared_qe.nosym = True
shared_qe.job = scf_job
shared_qe.pseudos = params.dft_pps
if params.tot_mag is not None:
shared_qe.tot_magnetization = params.tot_mag
#end if
if params.two_u:
ulist = [params.ulist[0]]
else:
ulist = params.ulist
#end if
shared_relax = shared_qe.copy()
# qmcs list to possibly bundle all DMC calculations
qmcs = []
scf_ks_energy = 0
for u in ulist:
if u != 0.:
shared_qe.hubbard_u = obj()
shared_relax.hubbard_u = obj()
for inum, i in enumerate(params.uatoms):
if params.two_u:
setattr(shared_qe.hubbard_u, i, params.ulist[inum])
else:
setattr(shared_qe.hubbard_u, i, u)
setattr(shared_relax.hubbard_u, i, params.relax_u)
#end if
#end for
#end if
scf_inf = None
scf_path = './scf-u-' + str(u) + '-inf'
if params.relax:
shared_relax = shared_qe.copy()
shared_relax['ion_dynamics'] = 'bfgs'
shared_relax['cell_dynamics'] = 'bfgs'
shared_relax['calculation'] = 'vc-relax'
shared_relax['conv_thr'] = 10e-6
shared_relax['forc_conv_thr'] = 0.001
shared_relax['input_DFT'] = params.relax_functional
shared_relax.nosym = True
coarse_relax = generate_pwscf(
identifier='scf',
path='./coarse_relax-u-' + str(params.relax_u) + '-inf',
electron_maxstep=params.electron_maxstep,
nogamma=True,
system=params.system,
kgrid=params.dft_grid,
**shared_relax)
sims.append(coarse_relax)
shared_relax['conv_thr'] = 10e-8
shared_relax['forc_conv_thr'] = 0.001
fine_relax = generate_pwscf(
identifier='scf',
path='./fine_relax-u-' + str(params.relax_u) + '-inf',
electron_maxstep=params.electron_maxstep,
nogamma=True,
system=params.system,
kgrid=params.dft_grid,
dependencies=(coarse_relax, 'structure'),
**shared_relax)
sims.append(fine_relax)
scf_inf = generate_pwscf(
identifier='scf',
path=scf_path,
electron_maxstep=params.electron_maxstep,
nogamma=True,
system=params.system,
#nosym = True,
kgrid=params.dft_grid,
calculation='scf',
dependencies=(fine_relax, 'structure'),
**shared_qe)
#pdb.set_trace()
else:
scf_inf = generate_pwscf(
identifier='scf',
path=scf_path,
electron_maxstep=params.electron_maxstep,
nogamma=True,
system=params.system,
#nosym = True,
kgrid=params.dft_grid,
calculation='scf',
**shared_qe)
sims.append(scf_inf)
if params.nscf_only is True:
params.tilings = []
params.vmc_opt = False
params.qmc = False
if params.print_ke_corr:
scf_dir = scf_inf.remdir
pwscf_output = scf_inf.input.control.outdir
datafile_xml = scf_dir + '/' + pwscf_output + '/pwscf.save/data-file-schema.xml'
if os.path.isfile(datafile_xml):
xmltree = minidom.parse(datafile_xml)
ks_energies = xmltree.getElementsByTagName("ks_energies")
eig_tot = 0
for kpt in ks_energies:
w = float(
kpt.getElementsByTagName('k_point')[0].getAttribute(
'weight'))
eigs = np.array(kpt.getElementsByTagName('eigenvalues')
[0].firstChild.nodeValue.split(' '),
dtype='f')
occs = np.array(
kpt.getElementsByTagName(
'occupations')[0].firstChild.nodeValue.split(' '),
dtype='f') #Non-integer occupations from DFT
eig_tot += w * sum(
eigs * occs) / 2 #Divide by two for both spins
#end for
print scf_dir, "e_tot eigenvalues in Ha ", eig_tot / 2
if scf_ks_energy == 0:
scf_ks_energy = eig_tot / 2
#end if
#end if
for tl_num, tl in enumerate(params.tilings):
for k in params.qmc_grids:
knum = k[1] * k[2] * k[0]
if params.relax:
relax = fine_relax #prim = scf_inf.load_analyzer_image().input_structure.copy()
print 'Make sure the previous run is loaded to results!'
#scf_inf.system.structure
else:
relax = scf_inf
prim = params.system.structure.copy()
#system = params.system
prim.clear_kpoints()
super = prim.tile(tl)
tl_np = np.array(tl)
tl_det = int(np.abs(np.linalg.det(tl_np)) + 0.001)
if params.natoms is not None:
natoms = tl_det * params.natoms
else:
natoms = len(super.elem)
#end if
if params.rcut:
rcut = params.rcut[tl_num]
else:
rcut = super.rwigner() - 0.00001
#end if
if params.tot_mag is not None and params.qp is False:
system = generate_physical_system(
structure=prim,
tiling=tl,
kgrid=k,
kshift=params.kshift, #(0, 0, 0),
net_charge=params.charge,
net_spin=params.tot_mag,
use_prim=False,
**params.system.valency)
elif params.qp is False:
system = generate_physical_system(
structure=prim,
tiling=tl,
kgrid=k,
shift=params.kshift, #(0, 0, 0),
net_charge=params.charge,
**params.system.valency)
else:
system = generate_physical_system(
structure=prim,
tiling=tl,
kgrid=k,
kshift=params.kshift, #(0, 0, 0),
use_prim=False,
**params.system.valency)
#end if
# DFT NSCF To Generate Wave Function At Specified K-points
if params.dft_grid == params.qmc_grids[0] and len(
params.qmc_grids) == 1:
params.nscf_skip = True
if not params.nscf_skip:
if params.relax:
nscf_dep = [(scf_inf, 'charge_density'),
(relax, 'structure')]
else:
nscf_dep = [(scf_inf, 'charge_density')]
#end if
nscf_path = './nscf-u-' + str(u) + '-' + str(
knum) + '-' + str(natoms)
p2q_path = nscf_path
nscf = generate_pwscf(
identifier='nscf',
path=nscf_path,
system=system,
#nosym = True,
nogamma=True,
dependencies=nscf_dep,
calculation='nscf',
**shared_qe)
sims.append(nscf)
if params.relax:
p2q_dep = [(nscf, 'orbitals'), (relax, 'structure')]
else:
p2q_dep = [(nscf, 'orbitals')]
#end if
else:
p2q_path = scf_path
if params.relax:
p2q_dep = [(scf_inf, 'orbitals'), (relax, 'structure')]
else:
p2q_dep = [(scf_inf, 'orbitals')]
#end if
#end if
# Convert DFT Wavefunction Into HDF5 File For QMCPACK
p2q = generate_pw2qmcpack(
identifier='p2q',
path=p2q_path,
job=p2q_job,
write_psir=False,
dependencies=p2q_dep,
)
sims.append(p2q)
if params.print_ke_corr and not params.nscf_skip:
import h5py
nscf_dir = nscf.remdir
pwscf_output = nscf.input.control.outdir
h5_file = nscf_dir + '/' + pwscf_output + '/pwscf.pwscf.h5'
#Assume equal weight for all k-points
eig_tot = 0
nkpts = 0
if os.path.isfile(h5_file):
f = h5py.File(h5_file, 'r+')
nelect = f['electrons']['number_of_electrons'][:]
for group_e in f['electrons'].keys():
if group_e.startswith('kpoint'):
nkpts += 1
f_temp = f['electrons'][group_e]
for group_k in f_temp.keys():
k_s_eig = [0]
w = 0
if group_k.startswith('spin_0'):
k_s_eig = f_temp[group_k][
'eigenvalues'][:]
k_s_eig = k_s_eig[0:nelect[
0]] #fixed occupations from the number of electrons
w = f_temp['weight'][0]
elif group_k.startswith('spin_1'):
k_s_eig = f_temp[group_k][
'eigenvalues'][:]
k_s_eig = k_s_eig[0:nelect[1]]
w = f_temp['weight'][0]
#pdb.set_trace()
#end if
eig_tot += sum(k_s_eig) * w / 2
#end for
#end if
#end fof
print nscf_dir, "e_tot eigenvalues in Ha", eig_tot / 2 - scf_ks_energy, "wigner " + str(
system['structure'].rwigner())
#end if
#end if
# DFT is complete, rest is QMC
# change here for other AFM atoms
system.rename(Co1='Co', Co2='Co', Co3='Co', folded=False)
system.rename(Ni1='Ni', Ni2='Ni', Ni3='Ni', folded=False)
# VMC Optimization
linopt1 = linear(
energy=0.0, # 0.95
unreweightedvariance=1.0,
reweightedvariance=0.0, # 0.05
timestep=0.3,
samples=samples,
walkers=walkers,
warmupsteps=10,
blocks=blocks,
steps=1,
substeps=10,
maxweight=1e9,
gpu=True,
minmethod='OneShiftOnly',
minwalkers=0.01,
usebuffer=True,
exp0=-6,
bigchange=10.0,
alloweddifference=1e-04,
stepsize=0.15,
nstabilizers=1,
)
# Quasiparticle calculation
if params.qp:
etot = system.particles.down_electron.count + system.particles.up_electron.count
upe = (etot + params.tot_mag * tl_det) / 2 - params.charge
downe = etot - params.charge - upe
system.particles.down_electron.count = downe
system.particles.up_electron.count = upe
#end if
# 1. VMC optimization is requested
# 2. VMC optimization is done on the first element uf ulist
# 3. VMC optimization is done on the first element of qmc_grids
if params.vmc_opt and u == params.ulist[
0] and k == params.qmc_grids[0]:
# VMC Variance minimization
opt_varmin = generate_qmcpack(
identifier='opt-varmin',
path='./opt-u-' + str(u) + '-' + str(natoms) +
'-varmin-' + str(params.j),
job=opt_job,
input_type='basic',
system=system,
spin_polarized=True, # jtk: needs this
meshfactor=1.0,
hybrid_rcut=params.hybrid_rcut,
hybrid_lmax=params.hybrid_lmax,
#spline_radius = params.spline_radius,
twistnum=params.twistnum,
#bconds = 'ppp',
pseudos=params.qmc_pps,
jastrows=[('J1', 'bspline', params.j, rcut),
('J2', 'bspline', params.j, rcut, 'init',
'zero')],
calculations=[
loop(max=6, qmc=linopt1),
loop(max=6, qmc=linopt1)
],
dependencies=(p2q, 'orbitals'))
sims.append(opt_varmin)
# QMC Optimization Parameters - Finer Sampling Set -- Energy Minimization
linopt2 = linopt1.copy()
linopt2.minwalkers = 0.5
linopt2.energy = 0.95
linopt2.unreweightedvariance = 0.0
linopt2.reweightedvariance = 0.05
linopt3 = linopt2.copy()
linopt3.minwalkers = 0.5
##
emin_dep = []
# Use preliminary optimization step made of two steps
if not params.nopre:
preopt_emin = generate_qmcpack(
identifier='opt-emin',
path='./preopt-u-' + str(u) + '-' + str(natoms) +
'-emin-' + str(params.j),
job=opt_job,
input_type='basic',
system=system,
spin_polarized=True, # jtk: needs this
meshfactor=params.meshfactor,
hybrid_rcut=params.hybrid_rcut,
hybrid_lmax=params.hybrid_lmax,
#spline_radius = params.spline_radius,
twistnum=params.twistnum,
#bconds = 'ppp',
pseudos=params.qmc_pps,
jastrows=[],
calculations=[loop(max=2, qmc=linopt2)],
dependencies=[(p2q, 'orbitals'),
(opt_varmin, 'jastrow')])
sims.append(preopt_emin)
emin_dep = preopt_emin
else:
emin_dep = opt_varmin
#end if
opt_emin = generate_qmcpack(
identifier='opt-emin',
path='./opt-u-' + str(u) + '-' + str(natoms) +
'-emin-' + str(params.j),
job=opt_job,
input_type='basic',
system=system,
spin_polarized=True, # jtk: needs this
meshfactor=params.meshfactor,
hybrid_rcut=params.hybrid_rcut,
hybrid_lmax=params.hybrid_lmax,
#spline_radius = params.spline_radius,
twistnum=params.twistnum,
#bconds = 'ppp',
pseudos=params.qmc_pps,
jastrows=[],
calculations=[
loop(max=4, qmc=linopt2),
loop(max=4, qmc=linopt3)
],
dependencies=[(p2q, 'orbitals'),
(emin_dep, 'jastrow')])
sims.append(opt_emin)
# 3-body jastrows
if params.j3:
j3_dep = []
linopt2.walkers = 16
linopt2.blocks = linopt2.blocks * 2
linopt2.samples = linopt1.samples * 2
j3_rcut = min(system['structure'].rwigner() - 0.0001,
4.0)
if not params.nopre:
preopt_emin_J3 = generate_qmcpack(
identifier='opt-emin-J3',
path='./preopt-u-' + str(u) + '-' +
str(natoms) + '-emin-J3-' + str(params.j),
job=opt_job,
input_type='basic',
system=system,
spin_polarized=True, # jtk: needs this
meshfactor=params.meshfactor,
hybrid_rcut=params.hybrid_rcut,
hybrid_lmax=params.hybrid_lmax,
#spline_radius = params.spline_radius,
twistnum=params.twistnum,
#bconds = 'ppp',
pseudos=params.qmc_pps,
jastrows=[('J3', 'polynomial', 3, 3, j3_rcut)],
calculations=[loop(max=2, qmc=linopt2)],
dependencies=[(p2q, 'orbitals'),
(opt_emin, 'jastrow')])
sims.append(preopt_emin_J3)
j3_dep = preopt_emin_J3
else:
j3_dep = opt_emin
#end if
opt_emin_J3 = generate_qmcpack(
identifier='opt-emin-J3',
path='./opt-u-' + str(u) + '-' + str(natoms) +
'-emin-J3-' + str(params.j),
job=opt_job,
input_type='basic',
system=system,
spin_polarized=True, # jtk: needs this
meshfactor=params.meshfactor,
hybrid_rcut=params.hybrid_rcut,
hybrid_lmax=params.hybrid_lmax,
#spline_radius = params.spline_radius,
twistnum=params.twistnum,
#bconds = 'ppp',
pseudos=params.qmc_pps,
jastrows=[('J3', 'polynomial', 3, 3, j3_rcut)],
calculations=[
loop(max=4, qmc=linopt2),
loop(max=4, qmc=linopt2)
],
dependencies=[(p2q, 'orbitals'),
(j3_dep, 'jastrow')])
sims.append(opt_emin_J3)
#end if
#end if
# VMC is complete, run DMC
dmc_nodes = np.round(
np.ceil(knum * dmcnodespertwist) / minnodes) * minnodes
dmc_job = Job(nodes=int(dmc_nodes),
hours=dmc_hours,
threads=qmc_threads,
app=qmcapp,
presub=qmc_presub)
# Add any future estimators here
from qmcpack_input import spindensity, skall, density
if params.density is not None:
est = [spindensity(grid=params.density)]
else:
est = None
#end if
#Initial VMC calculation to generate walkers for DMC
calculations = [
vmc(
warmupsteps=25,
blocks=vmcblocks,
steps=1,
stepsbetweensamples=1,
walkers=walkers,
timestep=vmcdt,
substeps=4,
samplesperthread=int(dmcwalkers /
(dmcnodespertwist * walkers)),
),
# dmc(
# warmupsteps =0,
# blocks =dmceqblocks/4,
# steps =5,
# timestep =0.04,
# nonlocalmoves=params.tmoves,
# ),
# dmc(
# warmupsteps =0,
# blocks =dmceqblocks/4,
# steps =5,
# timestep =0.02,
# nonlocalmoves=params.tmoves,
# ),
]
# Scan over timesteps
if params.timesteps is None:
params.timesteps = [0.01]
#end if
for t in params.timesteps:
calculations.append(
dmc(
warmupsteps=dmceqblocks / 2,
blocks=int(dmcblocks / (np.sqrt(t) * 10)),
steps=10,
timestep=t,
nonlocalmoves=params.tmoves,
))
#end for
# Prepanding name for the path
pathpre = 'dmc'
if params.vmc:
pathpre = 'vmc'
#end if
deps = [(p2q, 'orbitals')]
has_jastrow = False
# 3-body jastrow calculation with VMC or DMC
if params.j3:
has_jastrow = True
# add j3 dependenciies
if params.vmc_opt:
deps.append((opt_emin_J3, 'jastrow'))
#end if
if params.tmoves:
path = './' + pathpre + '-j3-tm-u-' + str(
u) + '-' + str(knum) + '-' + str(natoms)
else:
path = './' + pathpre + '-j3-nl-u-' + str(
u) + '-' + str(knum) + '-' + str(natoms)
#end if
#end if
# 2-body jastrow calculation with VMC or DMC
if params.j2:
has_jastrow = True
#add j2 dependencies
if params.vmc_opt:
deps.append((opt_emin, 'jastrow'))
#end if
if params.tmoves:
path = './' + pathpre + '-j2-tm-u-' + str(
u) + '-' + str(knum) + '-' + str(natoms)
else:
path = './' + pathpre + '-j2-nl-u-' + str(
u) + '-' + str(knum) + '-' + str(natoms)
#end if
#end if
# No jastrow calculation with VMC or DMC
if not has_jastrow:
if params.tmoves:
path = './' + pathpre + '-j0-tm-u-' + str(
u) + '-' + str(knum) + '-' + str(natoms)
else:
path = './' + pathpre + '-j0-nl-u-' + str(
u) + '-' + str(knum) + '-' + str(natoms)
#end if
#end if
det_format = 'new'
# Optical excitation
if params.excitation is not None:
path += '_' + params.excitation[
0] + '_' + params.excitation[1].replace(" ", "_")
det_format = 'old'
#end if
# Charged cell
if params.charge != 0:
path += '_' + str(params.charge)
#end if
if params.qmc:
qmc = generate_qmcpack(seed=params.seed,
skip_submit=params.qmc_skip_submit,
det_format=det_format,
identifier=pathpre,
path=path,
job=dmc_job,
input_type='basic',
system=system,
estimators=est,
meshfactor=params.meshfactor,
hybrid_rcut=params.hybrid_rcut,
hybrid_lmax=params.hybrid_lmax,
excitation=params.excitation,
pseudos=params.qmc_pps,
jastrows=[],
spin_polarized=True,
calculations=calculations,
dependencies=deps)
qmcs.append(qmc)
#end if
#end if
#end for
#end for
if params.bundle:
from bundle import bundle
qmcb = bundle(qmcs)
sims.append(qmcb)
else:
sims = sims + qmcs
return sims
linopt2 = linopt1.copy()
linopt2.samples = 16384
# QMC Optimization
opt = generate_qmcpack(
identifier = 'opt',
path = '.',
job = opt_job,
input_type = 'basic',
system = rocksalt_LiH,
twistnum = 0,
bconds = 'ppp',
pseudos = qmc_pps,
jastrows = [('J1','bspline',8),
('J2','bspline',8)],
calculations = [loop(max=4,qmc=linopt1),
loop(max=4,qmc=linopt2)],
dependencies = (p2q,'orbitals'),
)
pp = opt.input.get('pseudos')
pp.Li.set(
format = 'casino',
l_local = 's',
nrule = 2,
lmax = 2,
cutoff = 2.19,
)
pp.H.set(
format = 'casino',
l_local = 's',
nrule = 2,
beta = 0.025,
exp0 = -16,
bigchange = 15.0,
alloweddifference = 1e-4,
stepsize = 0.2,
stabilizerscale = 1.0,
nstabilizers = 3
)
linopt2 = linopt1.copy()
linopt2.samples = 20480 # opt w/ 20000 samples
linopt3 = linopt1.copy()
linopt3.samples = 40960 # opt w/ 40000 samples
opt_calcs = [loop(max=8,qmc=linopt1), # loops over opt's
loop(max=6,qmc=linopt2),
loop(max=4,qmc=linopt3)]
# specify DMC parameters
qmc_calcs = [
vmc(
walkers = 1,
warmupsteps = 30,
blocks = 20,
steps = 10,
substeps = 2,
timestep = .4,
samples = 2048
),
dmc(
size=8) # with 8 knots
],
# opt parameters (qmcpack)
perform_opt=True, # produce optimal jastrows
block_opt=False, # if true, ignores opt and qmc
skip_submit_opt=False, # if true, writes input files, does not run opt
opt_kpoint='L', # supercell k-point for the optimization
opt_calcs=[ # qmcpack input parameters for opt
loop(
max=4, # No. of loop iterations
qmc=linear( # linearized optimization method
energy=0.0, # cost function
unreweightedvariance=1.0, # is all unreweighted variance
reweightedvariance=0.0, # no energy or r.w. var.
timestep=0.5, # vmc timestep (1/Ha)
warmupsteps=100, # MC steps before data collected
samples=16000, # samples used for cost function
stepsbetweensamples=10, # steps between uncorr. samples
blocks=10, # ignore this
minwalkers=0.1, # and this
bigchange=15.0, # and this
alloweddifference=1e-4 # and this, for now
)),
loop(
max=4,
qmc=linear( # same as above, except
energy=0.5, # cost function
unreweightedvariance=0.0, # is 50/50 energy and r.w. var.
reweightedvariance=0.5,
timestep=0.5,
warmupsteps=100,
size = 8, # with 8 knots
rcut = 8.0), # and a radial cutoff of 8 bohr
],
# opt parameters (qmcpack)
perform_opt = True, # produce optimal jastrows
block_opt = False, # if true, ignores opt and qmc
skip_submit_opt = False, # if true, writes input files, does not run opt
opt_calcs = [ # qmcpack input parameters for opt
loop(max = 4, # No. of loop iterations
qmc = linear( # linearized optimization method
energy = 0.0, # cost function
unreweightedvariance = 1.0, # is all unreweighted variance
reweightedvariance = 0.0, # no energy or r.w. var.
timestep = 0.5, # vmc timestep (1/Ha)
warmupsteps = 100, # MC steps before data collected
samples = 16000,# samples used for cost function
stepsbetweensamples = 10, # steps between uncorr. samples
blocks = 10, # ignore this
minwalkers = 0.1,# and this
bigchange = 15.0,# and this
alloweddifference = 1e-4 # and this, for now
)
)
],
# qmc parameters (qmcpack)
block_qmc = False, # if true, ignores qmc
skip_submit_qmc = False, # if true, writes input file, does not run qmc
qmc_calcs = [ # qmcpack input parameters for qmc
vmc( # vmc parameters
timestep = 0.5, # vmc timestep (1/Ha)
warmupsteps = 100, # No. of MC steps before data is collected
linopt4.minwalkers = 0.85 # 0.5
linopt4.samples = linopt3.samples*4
#### optimize 12-body Jastrow
optJ12 = generate_qmcpack(
identifier = 'optJ12',
path = 'optJ12',
job = job(nodes=1, queue='debug', hours=0.5, constraint='knl', presub=scf_presub),
#job = job(nodes=1, threads=4),
system = system,
J2 = True, # 2-body B-spline Jastrow
J1_rcut = 5.0, # 4 Bohr cutoff for J1
J2_rcut = 5.0, # 7 Bohr cutoff for J2
pseudos = ['Be.ccECP.xml', 'H.ccECP.xml'],
calculations = [
loop(max=3, qmc=linopt1),
loop(max=3, qmc=linopt2),
loop(max=3, qmc=linopt3),
],
dependencies = orbdeps,
)
sims.append(optJ12)
##### optimize 3-body Jastrow
optJ123 = generate_qmcpack(
identifier = 'optJ123',
path = 'optJ123',
job = job(nodes=1, queue='debug', hours=0.5, constraint='knl', presub=scf_presub),
#job = job(nodes=1, threads=4),
system = system,
J3 = True, # 2-body B-spline Jastrow
beta = 0.025,
exp0 = -16,
bigchange = 15.0,
alloweddifference = 1e-4,
stepsize = 0.2,
stabilizerscale = 1.0,
nstabilizers = 3
)
linopt2 = linopt1.copy()
linopt2.samples = 20480 # opt w/ 20000 samples
linopt3 = linopt1.copy()
linopt3.samples = 40960 # opt w/ 40000 samples
opt_calcs = [loop(max=8,qmc=linopt1), # loops over opt's
loop(max=6,qmc=linopt2),
loop(max=4,qmc=linopt3)]
# specify DMC parameters
qmc_calcs = [
vmc(
walkers = 1,
warmupsteps = 30,
blocks = 20,
steps = 10,
substeps = 2,
timestep = .4,
samples = 2048
),
dmc(