def test_determine_convergence(self):
self.maxDiff = None
xs = [1, 2, 3, 4, 5, 6]
# a converging example:
ys = [4, 5, 6, 6, 6, 6]
self.assertEqual(
determine_convergence(xs, ys, name="name", tol=0.1, plots=False),
[True, 4, 6, 3, 6.0, 0.091269841269839724],
)
# another converging example
ys = [1 / 1, 1 / 2, 1 / 3, 1 / 4, 1 / 5, 1 / 6]
self.assertEqual(
determine_convergence(xs, ys, name="name", tol=0.3, plots=False),
[True, 3, 0.3333333333333333, 2, 0.0, -0.12813051146384496],
)
# a non converging example
ys = [4, 5, 6, 7, 8, 9]
self.assertEqual(
determine_convergence(xs, ys, name="name", tol=0.01, plots=False),
[False, numpy.inf, None, None, 14.607906815185412, None],
)
# another non converging example
ys = [4, 5, 4, 5, 4, 5]
self.assertEqual(
determine_convergence(xs, ys, name="name", tol=0.01, plots=False),
[False, numpy.inf, None, None, 11.368169147574115, None],
)
def find_conv_pars(self, tol=0.0001):
"""
find the pair of smallest values of ecuteps and nbands that give a gamma - gamma gap converged within tol
positive tol ensures the dirivative is smaller than tol
negative tol ensures the value is closer than -tol to the assymtotic limmit of a 'A + B / x ^ N' fit
"""
ecuteps_l = False
nbands_l = False
ecuteps_c = 0
nbands_c = 0
ecuteps_d = 0
nbands_d = 0
gap = None
y_conv = []
z_conv = []
y_conv_der = []
extrapolated = []
xs = self.get_var_range("nbands")
ys = self.get_var_range("ecuteps")
zd = self.get_data_array()
for z in zd:
print(z)
# print 'plot "'+self.name+'condat'+'"'
for x in xs:
zs = []
for y in ys:
try:
zs.append(zd[x][y])
except KeyError:
pass
conv_data = determine_convergence(ys, zs, name=self.name, tol=tol, extra="ecuteps at " + str(x))
extrapolated.append(conv_data[4])
if conv_data[0]:
y_conv.append(conv_data[1])
y_conv_der.append(conv_data[5])
z_conv.append(conv_data[2])
ecuteps_l = conv_data[0]
else:
y_conv.append(None)
z_conv.append(None)
if ecuteps_l:
conv_data = determine_convergence(xs, z_conv, name=self.name, tol=tol, extra="nbands")
if conv_data[0]:
nbands_l = conv_data[0]
nbands_c = conv_data[1]
gap = conv_data[2]
ecuteps_c = y_conv[conv_data[3]]
nbands_d = conv_data[5]
ecuteps_d = y_conv_der[conv_data[3]]
self.conv_res["control"].update({"ecuteps": ecuteps_l, "nbands": nbands_l})
self.conv_res["values"].update({"ecuteps": ecuteps_c, "nbands": nbands_c, "gap": gap})
self.conv_res["derivatives"].update({"ecuteps": ecuteps_d, "nbands": nbands_d})
return determine_convergence(
xs, extrapolated, name=self.name, tol=-0.05, extra="nbands at extrapolated ecuteps"
)
def find_conv_pars(self, tol=0.0001, silent=False):
"""
find the pair of smallest values of ecuteps and nbands that give a gamma - gamma gap converged within tol
positive tol ensures the dirivative is smaller than tol
negative tol ensures the value is closer than -tol to the assymtotic limmit of a 'A + B / x ^ N' fit
"""
plots = False if silent else True
ecuteps_l = False
nbands_l = False
ecuteps_c = 0
nbands_c = 0
ecuteps_d = 0
nbands_d = 0
gap = None
y_conv = []
z_conv = []
y_conv_der = []
extrapolated = []
xs = self.get_var_range('nbands')
ys = self.get_var_range('ecuteps')
zd = self.get_data_array()
# for z in zd:
# print(z)
# print 'plot "'+self.name+'condat'+'"'
for x in xs:
zs = []
for y in ys:
try:
zs.append(zd[x][y])
except KeyError as ex:
print(ex.message)
conv_data = determine_convergence(ys, zs, name=self.name, tol=tol, extra='ecuteps at '+str(x), plots=plots)
extrapolated.append(conv_data[4])
if conv_data[0]:
y_conv.append(conv_data[1])
y_conv_der.append(conv_data[5])
z_conv.append(conv_data[2])
ecuteps_l = conv_data[0]
else:
y_conv.append(None)
z_conv.append(None)
if ecuteps_l:
conv_data = determine_convergence(xs, z_conv, name=self.name, tol=tol, extra='nbands', plots=plots)
if conv_data[0]:
nbands_l = conv_data[0]
nbands_c = conv_data[1]
gap = conv_data[2]
ecuteps_c = y_conv[conv_data[3]]
nbands_d = conv_data[5]
ecuteps_d = y_conv_der[conv_data[3]]
self.conv_res['control'].update({'ecuteps': ecuteps_l, 'nbands': nbands_l})
self.conv_res['values'].update({'ecuteps': ecuteps_c, 'nbands': nbands_c, 'gap': gap})
self.conv_res['derivatives'].update({'ecuteps': ecuteps_d, 'nbands': nbands_d})
return determine_convergence(xs, extrapolated, name=self.name, tol=-0.05,
extra='nbands at extrapolated ecuteps', plots=plots)
def find_conv_pars(self, tol=0.0001):
"""
find the pair of smallest values of ecuteps and nbands that give a gamma - gamma gap converged within tol
positive tol ensures the dirivative is smaller than tol
negative tol ensures the value is closer than -tol to the assymtotic limmit of a 'A + B / x ^ N' fit
"""
ecuteps_l = False
nbands_l = False
ecuteps_c = 0
nbands_c = 0
ecuteps_d = 0
nbands_d = 0
gap = None
y_conv = []
z_conv = []
y_conv_der = []
extrapolated = []
xs = self.get_var_range('nbands')
ys = self.get_var_range('ecuteps')
zd = self.get_data_array()
for z in zd:
print(z)
# print 'plot "'+self.name+'condat'+'"'
for x in xs:
zs = []
for y in ys:
try:
zs.append(zd[x][y])
except KeyError:
pass
conv_data = determine_convergence(ys, zs, name=self.name, tol=tol, extra='ecuteps at '+str(x))
extrapolated.append(conv_data[4])
if conv_data[0]:
y_conv.append(conv_data[1])
y_conv_der.append(conv_data[5])
z_conv.append(conv_data[2])
ecuteps_l = conv_data[0]
else:
y_conv.append(None)
z_conv.append(None)
if ecuteps_l:
conv_data = determine_convergence(xs, z_conv, name=self.name, tol=tol, extra='nbands')
if conv_data[0]:
nbands_l = conv_data[0]
nbands_c = conv_data[1]
gap = conv_data[2]
ecuteps_c = y_conv[conv_data[3]]
nbands_d = conv_data[5]
ecuteps_d = y_conv_der[conv_data[3]]
self.conv_res['control'].update({'ecuteps': ecuteps_l, 'nbands': nbands_l})
self.conv_res['values'].update({'ecuteps': ecuteps_c, 'nbands': nbands_c, 'gap': gap})
self.conv_res['derivatives'].update({'ecuteps': ecuteps_d, 'nbands': nbands_d})
return determine_convergence(xs, extrapolated, name=self.name, tol=-0.05, extra='nbands at extrapolated ecuteps')
def test_determine_convergence(self):
self.maxDiff = None
xs = [1, 2, 3, 4, 5, 6]
# a converging example:
ys = [4, 5, 6, 6, 6, 6]
self.assertEqual(determine_convergence(xs, ys, name='name', tol=0.1, plots=False),
[True, 4, 6, 3, 6.0, 0.091269841269839724])
#self.assertTrue(os.path.isfile('name.fitdat'))
#self.assertTrue(os.path.isfile('plot-fits'))
# another converging example
ys = [1/1, 1/2, 1/3, 1/4, 1/5, 1/6]
self.assertEqual(determine_convergence(xs, ys, name='name', tol=0.3, plots=False),
[True, 3, 0.3333333333333333, 2, 0.0, -0.12813051146384496])
# a non converging example
ys = [4, 5, 6, 7, 8, 9]
self.assertEqual(determine_convergence(xs, ys, name='name', tol=0.01, plots=False),
[False, numpy.inf, None, None, 14.607906815185412, None])
# another non converging example
ys = [4, 5, 4, 5, 4, 5]
self.assertEqual(determine_convergence(xs, ys, name='name', tol=0.01, plots=False),
[False, numpy.inf, None, None, 11.368169147574115, None])
def find_conv_pars_scf(self, x_name, y_name, tol=0.0001):
xs = self.get_var_range(x_name)
ys = []
# print self.get_data_array_2d(x_name, y_name)
for x in xs:
ys.append(self.get_data_array_2d(x_name, y_name)[x])
conv_data = determine_convergence(xs, ys, name=self.name, tol=tol, extra=x_name)
# print conv_data, {x_name: conv_data[0]}, {x_name: conv_data[1]}, {x_name: conv_data[5]}
self.conv_res["control"].update({x_name: conv_data[0]})
self.conv_res["values"].update({x_name: conv_data[1]})
self.conv_res["derivatives"].update({x_name: conv_data[5]})
return conv_data
def find_conv_pars_scf(self, x_name, y_name, tol=0.0001):
xs = self.get_var_range(x_name)
ys = []
#print self.get_data_array_2d(x_name, y_name)
for x in xs:
ys.append(self.get_data_array_2d(x_name, y_name)[x])
conv_data = determine_convergence(xs, ys, name=self.name, tol=tol, extra=x_name)
#print conv_data, {x_name: conv_data[0]}, {x_name: conv_data[1]}, {x_name: conv_data[5]}
self.conv_res['control'].update({x_name: conv_data[0]})
self.conv_res['values'].update({x_name: conv_data[1]})
self.conv_res['derivatives'].update({x_name: conv_data[5]})
return conv_data
def test_convergence(self):
xs = sorted(self.df_data.keys())
#print xs
ys = []
for x in xs:
ys.append(self.df_data[x])
#print ys
for tol in [-10, -3.0, -1.0, -0.3, -0.1]:
test_res = determine_convergence(xs, ys, 'df'+str(abs(tol)), tol=tol, verbose=False, mode='extra_noise')
self.results.update({abs(tol): test_res[1]})
self.df_extra = test_res[4]
pprint.pprint(self.results)
self.pseudo.dojo_report.update({'hints': {'high': self.results[1.0], 'medium': self.results[3.0],
'low': self.results[10], 'based_on': 'delta_factor'}})
def test_convergence(self):
xs = sorted(self.df_data.keys())
#print xs
ys = []
for x in xs:
ys.append(self.df_data[x])
#print ys
for tol in [-10, -3.0, -1.0, -0.3, -0.1]:
test_res = determine_convergence(xs, ys, 'df'+str(abs(tol)), tol=tol, verbose=False, mode='extra_noise')
self.results.update({abs(tol): test_res[1]})
self.df_extra = test_res[4]
pprint.pprint(self.results)
self.pseudo.dojo_report.update({'hints': {'high': self.results[1.0], 'medium': self.results[3.0],
'low': self.results[10], 'based_on': 'delta_factor'}})
def find_conv_pars_scf(self, x_name, y_name, tol=0.0001, silent=False):
xs = self.get_var_range(x_name)
ys = []
# print self.get_data_array_2d(x_name, y_name)
for x in xs:
ys.append(self.get_data_array_2d(x_name, y_name)[x])
conv_data = determine_convergence(xs, ys, name=self.name, tol=tol, extra=x_name, plots=not silent)
# print conv_data, {x_name: conv_data[0]}, {x_name: conv_data[1]}, {x_name: conv_data[5]}
self.conv_res['control'].update({x_name: conv_data[0]})
factor = 1/eV_to_Ha if x_name == 'ecut' and self.code_interface.hartree_parameters else 1
self.conv_res['values'].update({x_name: conv_data[1]*factor})
print(conv_data[1])
self.conv_res['derivatives'].update({x_name: conv_data[5]})
return conv_data
def find_conv_pars_scf(self, x_name, y_name, tol=0.0001, silent=False):
xs = self.get_var_range(x_name)
ys = []
# print self.get_data_array_2d(x_name, y_name)
for x in xs:
ys.append(self.get_data_array_2d(x_name, y_name)[x])
conv_data = determine_convergence(xs,
ys,
name=self.name,
tol=tol,
extra=x_name,
plots=not silent)
# print conv_data, {x_name: conv_data[0]}, {x_name: conv_data[1]}, {x_name: conv_data[5]}
self.conv_res['control'].update({x_name: conv_data[0]})
factor = 1 / eV_to_Ha if x_name == 'ecut' and self.code_interface.hartree_parameters else 1
self.conv_res['values'].update({x_name: conv_data[1] * factor})
print(conv_data[1])
self.conv_res['derivatives'].update({x_name: conv_data[5]})
return conv_data
def conv_plots(self, query=None, silent=False, title=None):
"""
plot the convergence plots for all entries in query
"""
if query is None:
query = {}
for item in self.col.find(query):
if not silent:
print('System : ', item['system'])
print('Ps : ', item['ps'])
print('extra : ', item['extra_vars'])
print('gwresults : ', item['gw_results'])
try:
if not silent:
print('data : ', item['data_file'])
if True: # isinstance(item['data_file'], ObjectId):
f = self.gfs.get(item['data_file'])
data = f.read()
xx = []
yy = []
x = []
y = []
z = []
for line in data.splitlines():
if len(line) > 1 and 'data' not in line:
x.append(float(line.split()[0]))
y.append(float(line.split()[1]))
z.append(float(line.split()[2]))
# collect the gap vs ecuteps curve for calculating the interpolated ecuteps
if abs(float(line.split()[0]) - item['gw_results']['nbands']) < 20:
xx.append(float(line.split()[1]))
yy.append(float(line.split()[2]))
tol = 0.05
convres = determine_convergence(xs=xx, ys=yy, name='test', tol=-tol)
n = convres[3]
if n > 0:
dx = abs(xx[n] - xx[n-1])
dy = abs(yy[n] - yy[n-1])
print(n, dx, dy)
if convres[2] < convres[4]:
m = convres[4] - tol
f = (m - yy[n-1]) / dy
else:
m = convres[4] + tol
f = (yy[n-1] - m) / dy
print('extrapol: %s, border: %s, factor %s ' % (convres[4], m, f))
ecuteps_interpol = xx[n-1] + f * dx
else:
ecuteps_interpol = convres[1]
ecuteps_interpol = ecuteps_interpol if ecuteps_interpol < 500000 else None
item['gw_results']['ecuteps_interpol'] = ecuteps_interpol
print('interpolated ecuteps: %s' % ecuteps_interpol)
self.col.save(item)
if title is None:
t = item['system']
else:
t = title
try:
p = ConvTest(title=t, x=x, y=y, z=z)
p.show()
except ValueError:
print('bad data')
except (KeyError, NoFile):
print('no convergence plot in DataBase')
return p.return_fig_ax()
