def test_simstats():
import numpy as np
from testing import value_eq
from numerics import simstats
m,v,e,k = simstats(rstream)
m_ref = 0.507154277182
v_ref = 0.0840257344388
e_ref = 0.00949486225214
k_ref = 1.07291426596
assert(value_eq(float(m),m_ref))
assert(value_eq(float(v),v_ref))
assert(value_eq(float(e),e_ref))
assert(value_eq(float(k),k_ref))
m_ref = np.array(10*[m_ref])
v_ref = np.array(10*[v_ref])
e_ref = np.array(10*[e_ref])
k_ref = np.array(10*[k_ref])
m,v,e,k = simstats(rstream_wide)
assert(value_eq(m,m_ref))
assert(value_eq(v,v_ref))
assert(value_eq(e,e_ref))
assert(value_eq(k,k_ref))
def read_stat_h5(self,*files,equil=0,savefile=None):
save = False
if savefile is not None:
if os.path.exists(savefile):
vlog('\nLoading from save file {}'.format(savefile))
self.load(savefile)
vlog('Done',n=1,time=True)
return
else:
save = True
#end if
#end if
vlog('\nReading n(k) data from stat.h5 files',time=True)
k = []
nk = []
nke = []
if len(files)==1 and isinstance(files[0],(list,tuple)):
files = files[0]
#end if
for file in files:
if isinstance(file,StatFile):
stat = file
else:
vlog('Reading stat.h5 file',n=1,time=True)
stat = StatFile(file,observables=['momentum_distribution'])
#end if
vlog('Processing n(k) data from stat.h5 file',n=1,time=True)
vlog('filename = {}'.format(stat.filepath),n=2)
group = stat.observable_groups(self,single=True)
kpoints = np.array(group['kpoints'])
nofk = np.array(group['value'])
nk_mean,nk_var,nk_error,nk_kappa = simstats(nofk[equil:],dim=0)
k.extend(kpoints)
nk.extend(nk_mean)
nke.extend(nk_error)
#end for
vlog('Converting concatenated lists to arrays',n=1,time=True)
data = obj(
tot = obj(
k = np.array(k),
nk = np.array(nk),
nk_err = np.array(nke),
)
)
self.set_attribute('raw',data)
if save:
vlog('Saving to file {}'.format(savefile),n=1)
self.save(savefile)
#end if
vlog('stat.h5 file read finished',n=1,time=True)
def md_statistics(self, equil=None, autocorr=None):
import numpy as np
from numerics import simstats, simplestats
mds = obj()
for q, v in self.md_data.items():
if equil is not None:
v = v[equil:]
#end if
if autocorr is None:
mean, var, error, kappa = simstats(v)
else:
nv = len(v)
nb = int(np.floor(float(nv) / autocorr))
nexclude = nv - nb * autocorr
v = v[nexclude:]
v.shape = nb, autocorr
mean, error = simplestats(v.mean(axis=1))
#end if
mds[q] = mean, error
#end for
return mds
nm.append(nm_tmp)
#end for
nm = np.array(nm)
# Obtain dimensions of number matrices
nblocks, nstates, nstates = nm[0][0].shape
# Store stats of number matrix corresponding to single determinant with no jastrow, projected
# on MO basis
from numerics import simstats
m_mo, v_mo, e_mo, k_mo = simstats(nm, dim=2) # stats over blocks
# Perform "unitary" transform on each block's number matrix individually
# and store in nmqmcu (i.e., up component of number matrix prime)
# After the transformation, number matrix has been transformed from
# the MO basis to the AO basis
s = sp
nmqmc = np.empty((nKpoints, nSpin, nblocks, nAtomicWFC, nAtomicWFC),
dtype=complex)
for k in range(nKpoints):
for b in range(nblocks):
nmqmc[k][s][b] = kWeights[k] * np.matmul(
atomicProjections[k][s][:, :],
np.matmul(nm[k][0][b][:, :],
