def generate_A(pars, step_factor=1.0, lorentzian=False, **kwargs):
if lorentzian:
step_dir = 1j
else:
step_dir = -1
A = datadispenser.get_data(
dbname,
"A",
pars=pars,
complexion_step_direction=step_dir,
iter_count=0,
complexion_timestep=pars["complexion_timestep"] * step_factor,
**kwargs)[0]
return A
def main():
pars = parse()
apply_default_pars(pars, parinfo)
datadispenser.update_default_pars("timeevolved_insertion_mcmps", pars,
algorithm="MPS", t=3)
dbname = pars["database"]
if pars["debug"]:
warnings.filterwarnings('error')
datetime_str = datetime.datetime.strftime(datetime.datetime.now(),
'%Y-%m-%d_%H-%M-%S')
title_str = ('{}_{}'.format(filename, datetime_str))
logfilename = "logs/{}.log".format(title_str)
rootlogger = logging.getLogger()
set_filehandler(rootlogger, logfilename, pars)
# - Infoprint -
infostr = "\n{}\n".format("="*70)
infostr += "Running {} with the following parameters:".format(filename)
for k,v in sorted(pars.items()):
infostr += "\n%s = %s"%(k, v)
logging.info(infostr)
# Find the range of time and position for evaluating the
# expectations.
max_t_step = int(np.ceil(pars["max_t"]/pars["lorentzion_timestep"]))
t_steps = list(range(max_t_step+1))
ts = [pars["lorentzion_timestep"]*t_step for t_step in t_steps]
fid_t_file = datadir + "/fid_t_latest.npy"
llimit = pars["llimit"]
rlimit = pars["rlimit"]
dist = rlimit - llimit
if os.path.exists(fid_t_file):
os.remove(fid_t_file)
fid_t = np.empty((dist, 3, 0), dtype=np.complex_)
for t, t_step in zip(ts, t_steps):
t = np.around(t, 10)
logging.info("\nt: {}".format(t))
mcmps = datadispenser.get_data(
dbname, "timeevolved_insertion_mcmps", pars, t=t, algorithm="MPS"
)
N = mcmps.length()
eval_point = 0
w = mcmps.weights(eval_point)
ent = entanglement_entropy(w)
logging.info("Length: {}".format(N))
logging.info("Norm factors: {:.9e} & {:.9e}"
.format(mcmps.normfactor, mcmps.umps.normfactor))
umps = datadispenser.get_data(
dbname, "umps_groundstate", pars, algorithm="MPS"
)
conj_mps = type(mcmps)(umps)
# This identifier changing is some ad hoc crap that shouldn't be
# necessary.
conj_mps.change_identifier()
mcmps.change_identifier()
temp1 = np.array([mcmps.halfsystem_fidelity(conj_mps, i+1/2,
normalize=False)
for i in range(llimit, rlimit)])
temp2 = np.array([mcmps.window_fidelity(conj_mps, i, i+1,
normalize=False)
for i in range(llimit, rlimit)])
temp2 = np.reshape(temp2, (temp2.shape[0], 1))
temp = np.concatenate((temp1, temp2), axis=1)
temp = np.reshape(temp, (temp.shape[0], 3, 1))
fid_t = np.concatenate((fid_t, temp), axis=2)
np.save(fid_t_file, fid_t)
def main():
pars = parse()
apply_default_pars(pars, parinfo)
datadispenser.update_default_pars("umps_groundstate",
pars,
algorithm="MPS")
dbname = pars["database"]
if pars["debug"]:
warnings.filterwarnings('error')
datetime_str = datetime.datetime.strftime(datetime.datetime.now(),
'%Y-%m-%d_%H-%M-%S')
title_str = ('{}_{}'.format(filename, datetime_str))
logfilename = "logs/{}.log".format(title_str)
rootlogger = logging.getLogger()
set_filehandler(rootlogger, logfilename, pars)
# - Infoprint -
infostr = "\n{}\n".format("=" * 70)
infostr += "Running {} with the following parameters:".format(filename)
for k, v in sorted(pars.items()):
infostr += "\n%s = %s" % (k, v)
logging.info(infostr)
chi1 = pars["chi1"]
chi2 = pars["chi2"]
h1 = pars["h1"]
h2 = pars["h2"]
L = pars["L"]
do_separate = pars["do_separate"]
do_exact = pars["do_exact"]
umps = datadispenser.get_data(dbname,
"umps_groundstate",
pars,
algorithm="MPS",
mps_chis=range(1, chi1),
h_trans=h1)
mcmps1 = McMPS(umps, tensors=[umps.tensor.copy()], weightss=[])
logging.info("Correlation length 1: {}".format(
mcmps1.umps.correlation_length()))
umps = datadispenser.get_data(dbname,
"umps_groundstate",
pars,
algorithm="MPS",
mps_chis=range(1, chi2),
h_trans=h2)
mcmps2 = McMPS(umps, tensors=[umps.tensor.copy()], weightss=[])
logging.info("Correlation length 2: {}".format(
mcmps2.umps.correlation_length()))
if do_separate:
fids_separate = []
ul, ur = None, None
for i in range(L):
fid, ul, ur = mcmps1.window_fidelity_separate(mcmps2,
0,
i,
return_us=True,
initial_us=(ul, ur))
fids_separate.append(fid)
logging.info("{}, separate: {}".format(i, fid))
fids_separate = np.array(fids_separate)
with open(
"{}/fids_sep_latest_{}_{}_{}_{}_{}.p".format(
datadir, chi1, chi2, h1, h2, L), "wb") as f:
pickle.dump(fids_separate, f)
if do_exact:
fid0 = mcmps1.window_fidelity(mcmps2, 0, 0, log=True)
fids_exact = mcmps1.window_fidelity(mcmps2,
0,
L - 1,
upto=True,
log=True)
fids_exact = np.concatenate(([fid0], fids_exact))
with open(
"{}/fids_exact_latest_{}_{}_{}_{}_{}.p".format(
datadir, chi1, chi2, h1, h2, L), "wb") as f:
pickle.dump(fids_exact, f)
logfilename = "logs/{}.log".format(title_str)
rootlogger = logging.getLogger()
set_filehandler(rootlogger, logfilename, pars)
# - Infoprint -
infostr = "\n{}\n".format("=" * 70)
infostr += "Running {} with the following parameters:".format(filename)
for k, v in sorted(pars.items()):
infostr += "\n%s = %s" % (k, v)
logging.info(infostr)
dbname = pars["database"]
for it in pars["iters"]:
logging.info("\nIteration {}".format(it))
res = datadispenser.get_data(dbname, "A", pars, iter_count=it)
A, log_fact = res[0], res[1]
if pars["print_spectra"]:
es = get_A_spectrum(A)
msg = "Spectrum of A:\n{}".format(es[:30])
logging.info(msg)
if pars["print_free_energy"]:
f = get_free_energy(A, log_fact, pars, it)
exact_f = modeldata.get_free_energy(pars)
f_error = np.abs(f - exact_f) / exact_f
msg = ("Free energy per site: {} ({}, off by {:.4e})".format(
f, exact_f, f_error))
logging.info(msg)
def main():
pars = parse()
apply_default_pars(pars, parinfo)
datadispenser.update_default_pars("timeevolved_insertion_mcmps",
pars,
algorithm="MPS",
t=5)
datadispenser.update_default_pars("A",
pars,
iter_count=0,
complexion_step_direction=-1,
algorithm="TNR",
model="complexion_" + pars["model"],
complexion_spacestep=1,
complexion_timestep=1)
dbname = pars["database"]
if pars["debug"]:
warnings.filterwarnings('error')
datetime_str = datetime.datetime.strftime(datetime.datetime.now(),
'%Y-%m-%d_%H-%M-%S')
title_str = ('{}_{}'.format(filename, datetime_str))
logfilename = "logs/{}.log".format(title_str)
rootlogger = logging.getLogger()
set_filehandler(rootlogger, logfilename, pars)
# - Infoprint -
infostr = "\n{}\n".format("=" * 70)
infostr += "Running {} with the following parameters:".format(filename)
for k, v in sorted(pars.items()):
infostr += "\n%s = %s" % (k, v)
logging.info(infostr)
# Find the range of time and position for evaluating the
# expectations.
poses = list(range(-pars["max_x"], pars["max_x"] + 1))
max_t_step = int(np.ceil(pars["max_t"] / pars["lorentzion_timestep"]))
t_steps = list(range(max_t_step + 1))
ts = [pars["lorentzion_timestep"] * t_step for t_step in t_steps]
mcmps = datadispenser.get_data(dbname,
"timeevolved_insertion_mcmps",
pars,
t=0,
algorithm="MPS")
if mcmps.tensortype() == Tensor:
pars["symmetry_tensors"] = False
else:
pars["symmetry_tensors"] = True
op_late = get_operator_insertion(pars["insertion_late"])
op_late = mcmps.tensortype().from_ndarray(op_late)
# Evaluate correlators for various times.
data_filename = None
expectations = None
for t, t_step in zip(ts, t_steps):
t = np.around(t, 10)
logging.info("\nt: {}".format(t))
mcmps = datadispenser.get_data(dbname,
"timeevolved_insertion_mcmps",
pars,
t=t,
algorithm="MPS")
N = mcmps.length()
eval_point = 0
w = mcmps.weights(eval_point)
ent = entanglement_entropy(w)
logging.info("Length: {}".format(N))
logging.info("Norm factors: {:.9e} & {:.9e}".format(
mcmps.normfactor, mcmps.umps.normfactor))
logging.info("Entropy at {}: {}".format(eval_point, ent))
logging.info("Spectrum at {}:".format(eval_point))
logging.info(w)
old_data_filename = data_filename
expectations, data_filename, = update_expectations(
mcmps, op_late, t, expectations, poses, pars)
if old_data_filename is not None:
os.remove(old_data_filename)