def setUp(self):
self.wf = HarmOsc1D(nelec=1, ndim=1)
self.sampler = Metropolis(nwalkers=1000,
nstep=1000,
step_size=3,
nelec=1,
ndim=1,
domain={
'min': -2,
'max': 2
})
self.optimizer = Minimize(method='bfgs', maxiter=25, tol=1E-4)
def nuclear_potential(self, pos):
return (0.5 * pos**2).flatten()
def electronic_potential(self, pos):
return 0
if __name__ == "__main__":
wf = HarmOsc1D(nelec=1, ndim=1)
sampler = Metropolis(nwalkers=1000,
nstep=1000,
step_size=3,
nelec=1,
ndim=1,
domain={
'min': -2,
'max': 2
})
# sampler = PYMC3(nwalkers=1000,ndim=1)
optimizer = Minimize(method='bfgs', maxiter=25, tol=1E-4)
# optimizer = Swarm( maxiter=25)
# VMC solver
vmc = VMC(wf=wf, sampler=sampler, optimizer=optimizer)
# single point
opt_param = [0.5]
pos, e, s = vmc.single_point(opt_param)
# def __init__(self,model,nelec,ndim):
# NeuralWF.__init__(self, model, nelec, ndim)
# def nuclear_potential(self,pos):
# return torch.sum(0.5*pos**2,1)
# def electronic_potential(self,pos):
# return 0
# wf = HarmOsc3D(model=WaveNet,nelec=1, ndim=3)
wf = NEURAL_PYSCF_WF(atom='O 0 0 0; H 0 1 0; H 0 0 1',
basis='dzp',
active_space=(2, 2))
sampler = Metropolis(nwalkers=64,
nstep=10,
step_size=3,
nelec=wf.nelec,
ndim=3,
domain={
'min': -5,
'max': 5
})
nn = NN4PYSCF(wf=wf, sampler=sampler)
pos = nn.sample()
dataset = QMC_DataSet(pos)
dataloader = DataLoader(dataset, batch_size=nn.batchsize)
qmc_loss = QMCLoss(nn.wf, method='variance')