def test_constructor():
options = {
'verbosity': 0,
'get_sha1': False,
'qmc': {
'timestep': 0.01,
'num_steps': 10,
'blocks': 10,
'rng_seed': 8,
},
'estimates': {
'mixed': {
'energy_eval_freq': 1
}
}
}
model = {
'name': "UEG",
'rs': 2.44,
'ecut': 4,
'nup': 7,
'ndown': 7,
}
system = UEG(model)
trial = HartreeFock(system, True, {})
comm = MPI.COMM_WORLD
afqmc = AFQMC(comm=comm, options=options, system=system, trial=trial)
afqmc.finalise(verbose=0)
assert afqmc.trial.energy.real == pytest.approx(1.7796083856572522)
def get_driver(options, comm):
qmc_opts = get_input_value(options,
'qmc',
default={},
alias=['qmc_options'])
beta = get_input_value(qmc_opts, 'beta', default=None)
verbosity = options.get('verbosity', 1)
if beta is not None:
afqmc = ThermalAFQMC(comm,
options=options,
parallel=comm.size > 1,
verbose=verbosity)
else:
afqmc = AFQMC(comm,
options=options,
parallel=comm.size > 1,
verbose=verbosity)
return afqmc
def test_generic_single_det():
nmo = 11
nelec = (3, 3)
options = {
'verbosity': 0,
'qmc': {
'timestep': 0.005,
'num_steps': 10,
'blocks': 10,
'rng_seed': 8,
},
'trial': {
'name': 'hartree_fock'
},
'estimator': {
'back_propagated': {
'tau_bp': 0.025,
'one_rdm': True
},
'mixed': {
'energy_eval_freq': 1
}
}
}
numpy.random.seed(7)
h1e, chol, enuc, eri = generate_hamiltonian(nmo, nelec, cplx=False)
sys_opts = {'sparse': True}
sys = Generic(nelec=nelec, h1e=h1e, chol=chol, ecore=enuc, inputs=sys_opts)
comm = MPI.COMM_WORLD
afqmc = AFQMC(comm=comm, system=sys, options=options)
afqmc.run(comm=comm, verbose=0)
afqmc.finalise(verbose=0)
afqmc.estimators.estimators['mixed'].update(afqmc.system, afqmc.qmc,
afqmc.trial, afqmc.psi, 0)
enum = afqmc.estimators.estimators['mixed'].names
numer = afqmc.estimators.estimators['mixed'].estimates[enum.enumer]
denom = afqmc.estimators.estimators['mixed'].estimates[enum.edenom]
weight = afqmc.estimators.estimators['mixed'].estimates[enum.weight]
assert numer.real == pytest.approx(3.8763193646854273)
data = extract_mixed_estimates('estimates.0.h5')
assert numpy.mean(
data.ETotal.values[:-1].real) == pytest.approx(1.5485077038208)
rdm = extract_rdm('estimates.0.h5')
assert rdm[0, 0].trace() == pytest.approx(nelec[0])
assert rdm[0, 1].trace() == pytest.approx(nelec[1])
assert rdm[11, 0, 1, 3].real == pytest.approx(-0.121883381144845)
def test_hubbard():
options = {
'verbosity': 0,
'get_sha1': False,
'qmc': {
'timestep': 0.01,
'num_steps': 10,
'blocks': 10,
'rng_seed': 8,
},
'model': {
'name': "Hubbard",
'nx': 4,
'ny': 4,
'nup': 7,
"U": 4,
'ndown': 7,
},
'trial': {
'name': 'UHF'
},
'estimates': {
'mixed': {
'energy_eval_freq': 1
}
},
'propagator': {
'hubbard_stratonovich': 'discrete'
}
}
comm = MPI.COMM_WORLD
afqmc = AFQMC(comm=comm, options=options)
afqmc.run(comm=comm, verbose=0)
afqmc.finalise(verbose=0)
afqmc.estimators.estimators['mixed'].update(afqmc.system, afqmc.qmc,
afqmc.trial, afqmc.psi, 0)
enum = afqmc.estimators.estimators['mixed'].names
numer = afqmc.estimators.estimators['mixed'].estimates[enum.enumer]
denom = afqmc.estimators.estimators['mixed'].estimates[enum.edenom]
weight = afqmc.estimators.estimators['mixed'].estimates[enum.weight]
assert numer.real == pytest.approx(-152.68468568462666)
data = extract_mixed_estimates('estimates.0.h5')
# old code seemed to ommit last value. Discard to avoid updating benchmark.
assert numpy.mean(
data.ETotal.values[:-1]) == pytest.approx(-14.974806533852874)
def test_hubbard_complex():
options = {
'verbosity': 0,
'get_sha1': False,
'qmc': {
'timestep': 0.01,
'num_steps': 10,
'blocks': 10,
'rng_seed': 8,
},
'model': {
'name': "Hubbard",
'nx': 4,
'ny': 4,
'nup': 7,
"U": 4,
'ndown': 7,
},
'trial': {
'name': 'UHF'
},
'estimates': {
'mixed': {
'energy_eval_freq': 1
}
},
'propagator': {
'hubbard_stratonovich': 'continuous'
}
}
comm = MPI.COMM_WORLD
afqmc = AFQMC(comm=comm, options=options)
afqmc.run(comm=comm, verbose=0)
afqmc.finalise(verbose=0)
afqmc.estimators.estimators['mixed'].update(afqmc.system, afqmc.qmc,
afqmc.trial, afqmc.psi, 0)
enum = afqmc.estimators.estimators['mixed'].names
numer = afqmc.estimators.estimators['mixed'].estimates[enum.enumer]
denom = afqmc.estimators.estimators['mixed'].estimates[enum.edenom]
weight = afqmc.estimators.estimators['mixed'].estimates[enum.weight]
assert numer == pytest.approx(-152.91937839611)
data = extract_mixed_estimates('estimates.0.h5')
assert numpy.mean(
data.ETotal.values[:-1].real) == pytest.approx(-15.14323385684513)
def test_ueg():
options = {
'verbosity': 0,
'get_sha1': False,
'qmc': {
'timestep': 0.01,
'num_steps': 10,
'blocks': 10,
'rng_seed': 8,
},
'model': {
'name': "UEG",
'rs': 2.44,
'ecut': 4,
'nup': 7,
'ndown': 7,
},
'estimates': {
'mixed': {
'energy_eval_freq': 1
}
},
'trial': {
'name': 'hartree_fock'
}
}
comm = MPI.COMM_WORLD
# FDM: Fix cython issue.
afqmc = AFQMC(comm=comm, options=options)
afqmc.run(comm=comm, verbose=0)
afqmc.finalise(verbose=0)
afqmc.estimators.estimators['mixed'].update(afqmc.system, afqmc.qmc,
afqmc.trial, afqmc.psi, 0)
enum = afqmc.estimators.estimators['mixed'].names
numer = afqmc.estimators.estimators['mixed'].estimates[enum.enumer]
assert numer == pytest.approx(15.5272838037998)
denom = afqmc.estimators.estimators['mixed'].estimates[enum.edenom]
assert denom == pytest.approx(10)
weight = afqmc.estimators.estimators['mixed'].estimates[enum.uweight]
assert weight == pytest.approx(9.76035750882054)
ehy = afqmc.psi.walkers[0].hybrid_energy
assert ehy.real == pytest.approx(2.41659997842609)
assert ehy.imag == pytest.approx(-0.395817411848255)
def test_generic():
nmo = 11
nelec = (3, 3)
options = {
'verbosity': 0,
'get_sha1': False,
'qmc': {
'timestep': 0.005,
'steps': 10,
'blocks': 10,
'rng_seed': 8,
},
'estimates': {
'mixed': {
'energy_eval_freq': 1
}
},
'trial': {
'name': 'MultiSlater'
}
}
numpy.random.seed(7)
h1e, chol, enuc, eri = generate_hamiltonian(nmo, nelec, cplx=False)
sys = Generic(nelec=nelec, h1e=h1e, chol=chol, ecore=enuc)
comm = MPI.COMM_WORLD
afqmc = AFQMC(comm=comm, system=sys, options=options)
afqmc.run(comm=comm, verbose=0)
afqmc.finalise(verbose=0)
afqmc.estimators.estimators['mixed'].update(afqmc.system, afqmc.qmc,
afqmc.trial, afqmc.psi, 0)
enum = afqmc.estimators.estimators['mixed'].names
numer = afqmc.estimators.estimators['mixed'].estimates[enum.enumer]
denom = afqmc.estimators.estimators['mixed'].estimates[enum.edenom]
weight = afqmc.estimators.estimators['mixed'].estimates[enum.weight]
assert numer.real == pytest.approx(3.8763193646854273)
data = extract_mixed_estimates('estimates.0.h5')
assert numpy.mean(
data.ETotal.values[:-1].real) == pytest.approx(1.5485077038208)
def setup_parallel(options, comm=None, verbose=False):
"""Wrapper routine for initialising simulation
Parameters
----------
options : dict
Input options.
comm : MPI communicator
MPI communicator object.
verbose : bool
If true print out set up information.
Returns
-------
afqmc : :class:`pauxy.afqmc.CPMC`
CPMC driver.
"""
if comm.Get_rank() == 0:
afqmc = AFQMC(options.get('model'),
options.get('qmc_options'),
options.get('estimates'),
options.get('trial_wavefunction'),
options.get('propagator', {}),
parallel=True,
verbose=verbose)
else:
afqmc = None
afqmc = comm.bcast(afqmc, root=0)
afqmc.init_time = time.time()
if afqmc.trial.error:
warnings.warn('Error in constructing trial wavefunction. Exiting')
sys.exit()
afqmc.rank = comm.Get_rank()
afqmc.nprocs = comm.Get_size()
afqmc.root = afqmc.rank == 0
# We can't serialise '_io.BufferWriter' object, so just delay initialisation
# of estimators object to after MPI communication.
# Simpler to just ensure a fixed number of walkers per core.
afqmc.qmc.nwalkers = int(afqmc.qmc.nwalkers / afqmc.nprocs)
if afqmc.qmc.nwalkers == 0:
# This should occur on all processors so we don't need to worry about
# race conditions / mpi4py hanging.
if afqmc.root:
warnings.warn('Not enough walkers for selected core count. There '
'must be at least one walker per core set in the '
'input file. Exiting.')
sys.exit()
afqmc.estimators = (Estimators(options.get('estimates'), afqmc.root,
afqmc.qmc, afqmc.system, afqmc.trial,
afqmc.propagators.BT_BP))
afqmc.psi = Walkers(afqmc.system, afqmc.trial, afqmc.qmc.nwalkers,
afqmc.estimators.nprop_tot, afqmc.estimators.nbp)
if comm.Get_rank() == 0:
json.encoder.FLOAT_REPR = lambda o: format(o, '.6f')
json_string = json.dumps(serialise(afqmc, verbose=1),
sort_keys=False,
indent=4)
afqmc.estimators.h5f.create_dataset('metadata',
data=numpy.array([json_string],
dtype=object),
dtype=h5py.special_dtype(vlen=str))
return afqmc