def chi_list(chi_max, dchi=20, nsweeps=20, verbose=0):
warnings.warn("Deprecated: moved `chi_list` to `tenpy.algorithms.dmrg.chi_list`.",
category=FutureWarning,
stacklevel=2)
from tenpy.algorithms import dmrg
chi_list = dmrg.chi_list(chi_max, dchi, nsweeps)
if verbose:
import pprint
print("chi_list = ")
pprint.pprint(chi_list)
return chi_list
def test_chi_list():
assert dmrg.chi_list(3) == {0: 3}
assert dmrg.chi_list(12, 12, 5) == {0: 12}
assert dmrg.chi_list(24, 12, 5) == {0: 12, 5: 24}
assert dmrg.chi_list(27, 12, 5) == {0: 12, 5: 24, 10: 27}
def setup_executable(mod,
run_defaults,
identifier_list=None,
only_list_supplied=False):
"""Read command line arguments and turn into useable dicts.
Uses default values defined at:
- model class for model_par
- here for sim_par
- executable file for run_par
Alternatively, a model_defaults dictionary and identifier_list can be supplied without the model
NB: for setup_executable to work with a model class, the model class needs to define two things:
- defaults, a static (class level) dictionary with (key, value) pairs that have the name
of the parameter (as string) as key, and the default value as value.
- identifier, a static (class level) list or other iterable with the names of the parameters
to be used in filename identifiers.
Args:
mod (model | dict): Model class (or instance) OR a dictionary containing model defaults
run_defaults (dict): default values for executable file parameters
identifier_list (iterable, optional) | Used only if mod is a dict. Contains the identifier
variables
Returns:
model_par, sim_par, run_par (dicts) : containing all parameters.
args | namespace with raw arguments for some backwards compatibility with executables.
"""
warnings.warn(
"Attention: `setup_executable` was developed for a previous version of tenpy and not all options may be operational.",
category=FutureWarning,
stacklevel=2)
parser = argparse.ArgumentParser()
# These deal with backwards compatibility (supplying a model)
if type(
mod
) != dict and identifier_list == None: # Assume we've been given a model class
try:
model_defaults = mod.defaults
identifier_list = mod.identifier
except AttributeError as err:
print(
"Cannot get model defaults and identifer list from mod. Is mod a class/instance?"
)
print(err)
raise AttributeError
elif type(mod) == dict and hasattr(identifier_list, '__iter__'):
model_defaults = mod
else:
raise ValueError(
"If model_par are supplied as dict, identifier_list should be provided."
)
# The model_par bit (for all model parameters)
for label, value in model_defaults.items():
if type(value
) == bool: # For boolean defaults, we want a true/false flag
if value:
parser.add_argument('-' + label, action='store_false')
else:
parser.add_argument('-' + label, action='store_true')
else: # For non-boolean defaults, take the type of the default as type for the cmdline var
parser.add_argument('-' + label, type=type(value), default=value)
# The run_par bit (for executable-level parameters). These are defined in the executable file
# but need to be included for argparse to work correctly.
for label, value in run_defaults.items():
if type(value
) == bool: # For boolean defaults, we want a true/false flag
if value:
parser.add_argument('-' + label, action='store_false')
else:
parser.add_argument('-' + label, action='store_true')
else: # For non-boolean defaults, take the type of the default as type for the cmdline var
print('Adding argument', label)
parser.add_argument('-' + label, type=type(value), default=value)
# The following parameters are run-time but so general they're defined here
parser.add_argument('-ncores', type=int, default=1)
parser.add_argument('-dir', type=str, default=None)
parser.add_argument(
'-plots', action='store_true') # Generic flag to activate plotting
parser.add_argument('-seed', default=None) # For anything random
# The sim_par bit (for DMRG-related parameters). These don't vary, so we'll just define here.
if not 'active_sites' in run_defaults:
parser.add_argument('-active_sites', type=int, default=2)
parser.add_argument('-chi', type=int, default=100)
parser.add_argument('-dchi', type=int,
default=20) # Step size for chi ramp
parser.add_argument('-dsweeps', type=int,
default=20) # Number of sweeps for chi step
parser.add_argument('-nqramp', type=int,
default=0) # Number of insertions of qramp operator
parser.add_argument(
'-last_qramp', type=int,
default=50) # Last sweep at which an operator is inserted
parser.add_argument('-qramp_op', type=str,
default=None) # Operator to insert during ramp events
parser.add_argument(
'-qramp_site',
type=str,
default='0',
help=
'The site index where ramp operators should be inserted, or "R" for random sites'
) # Site on which to insert the ramp operator
parser.add_argument(
'-qramp_move_left', action='store_true'
) # flag indicating the operators should be inserted while moving left
if run_defaults.get('min_sweeps') is None:
parser.add_argument('-min_sweeps', type=int, default=30)
if run_defaults.get('max_sweeps') is None:
parser.add_argument('-max_sweeps', type=int, default=1000)
if run_defaults.get('N_sweeps_check') is None:
parser.add_argument('-N_sweeps_check', type=int, default=10)
# inputs for mixer parameters, as per https://tenpy.readthedocs.io/en/latest/reference/tenpy.algorithms.dmrg.Mixer.html#cfg-config-Mixer
parser.add_argument('-mixer', action='store_true') # To activate mixer
parser.add_argument('-mix_str', type=float, default=1.e-3)
parser.add_argument('-mix_dec', type=float, default=1.5)
parser.add_argument('-mix_len', type=int, default=80)
# control of tolerances:
if run_defaults.get('max_E_err') is None:
parser.add_argument('-max_E_err', type=float,
default=1e-8) # DMRG Error tolerance
if run_defaults.get('max_S_err') is None:
parser.add_argument('-max_S_err', type=float,
default=1e-5) # DMRG Entanglement tolerance
# DMRG norm tolerance: https://tenpy.readthedocs.io/en/latest/reference/tenpy.algorithms.dmrg.DMRGEngine.html#cfg-option-DMRGEngine.norm_tol
if not 'norm_tol' in run_defaults:
parser.add_argument(
'-norm_tol', type=float, default=1e-5
) # After the DMRG run, update the environment with at most `norm_tol_iter` sweeps until ``np.linalg.norm(psi.norm_err()) < norm_tol``.
if not 'norm_tol_iter' in run_defaults:
parser.add_argument('-norm_tol_iter', type=float, default=5.0)
#Perform at most `norm_tol_iter`*`update_env` sweeps to converge the norm error below `norm_tol`
# parameters controlling sweeps to reconstruct the environment
if run_defaults.get('start_env') is None:
parser.add_argument('-start_env', type=int, default=0)
if run_defaults.get('update_env') is None:
parser.add_argument('-update_env', type=int)
# Now parse and turn into manageable dicts.
args = parser.parse_args()
par_dict = vars(args) # Turns args (='Namespace' object) into dict.
model_par = {}
for label in model_defaults.keys(
): # Select the model-relevant parts of par_dict
model_par[label] = par_dict[label]
run_par = {}
for label in run_defaults.keys(
): # Select the executable-relevant parts of par_dict
run_par[label] = par_dict[label]
try:
from ..algorithms import dmrg
sim_par = {
'active_sites':
args.active_sites,
'chi_list':
dmrg.chi_list(args.chi, args.dchi, args.dsweeps),
'qramp_op':
args.qramp_op,
'qramp_list':
qramp_list(args.nqramp, args.last_qramp, args.qramp_op,
args.qramp_site, not args.qramp_move_left),
'min_sweeps':
args.min_sweeps,
'max_sweeps':
args.max_sweeps,
'N_sweeps_check':
args.N_sweeps_check,
'start_env':
args.start_env,
'max_E_err':
args.max_E_err,
'max_S_err':
args.max_S_err,
'norm_tol':
args.norm_tol,
'norm_tol_iter':
args.norm_tol_iter,
'verbose':
args.verbose, # Take this from the model
'lanczos_params': {
'N_min': 2,
'N_max': 40,
'E_tol': 10**(-12)
}
}
if (args.update_env is None):
sim_par['update_env'] = args.N_sweeps_check // 2
else:
sim_par['update_env'] = args.update_env
except AttributeError as err:
print(
'sim_par parsing has failed, most likely because model does not define verbose parameter.'
)
print(err)
raise AttributeError
if args.mixer:
sim_par['mixer'] = True
sim_par['mixer_params'] = {
'amplitude': args.mix_str,
'decay': args.mix_dec,
'disable_after': args.mix_len
}
# Having set up all dictionaries, we can now do some other setting up
omp_set_nthreads(args.ncores)
if not args.dir == None:
os.chdir(args.dir)
import matplotlib
matplotlib.rcParams["savefig.directory"] = os.chdir(os.getcwd())
# Build the identifier based on model-defined and general parameters
identifier = "chi_{}_seed_{}_".format(
args.chi, args.seed) # Only use seed if supplied?
for varname in identifier_list:
if 'conserve' in varname:
shortened = varname.replace('conserve', 'cons').replace(
'number', 'num').replace('charge', 'ch').replace('spin', 'S')
identifier += shortened + "_"
else:
if (only_list_supplied):
if model_par[varname] != model_defaults[varname]:
identifier += varname + "_" + str(model_par[varname]) + "_"
else:
if model_par[
varname] != 0: # Parameters that are 0 are ignored. Only want supplied?
identifier += varname + "_" + str(model_par[varname]) + "_"
if args.mixer:
identifier += 'mix_({},{},{})'.format(args.mix_str, args.mix_dec,
args.mix_len)
if identifier[-1] == "_":
identifier = identifier[:-1]
# Attempt to shorten the identifier
identifier = identifier.replace('periodic',
'inf').replace('finite',
'fin').replace('.0_', '_')
identifier = identifier.replace('flux_p',
'p').replace('flux_q', 'q').replace(
'phi_ext_mode', 'pe-mode')
if len(identifier) >= 144:
print(
"Warning: identifier has a length longer than max filename on encrypted Ubuntu! Try argument 'only_list_supplied'"
)
run_par.update({
'ncores': args.ncores,
'dir': args.dir,
'plots': args.plots,
'identifier': identifier,
'seed': args.seed,
})
return model_par, sim_par, run_par, args
