def dump_output(self, *print_properties):
""" Print output based on level of detail required
The default detail level (low) is the integrator's calc's update
property for each named particle array.
The higher detail level dumps all particle array properties.
"""
fname = self.fname + '_'
props = {}
cell_size = self.particles.cell_manager.cell_size
for pa in self.particles.arrays:
name = pa.name
_fname = os.path.join(self.output_directory,
fname + name + '_' + str(self.t) + '.npz')
if self.detailed_output:
savez(_fname, dt=self.dt, **pa.properties)
else:
for prop in print_properties:
props[prop] = pa.get(prop)
savez(_fname,
dt=self.dt,
cell_size=cell_size,
np=pa.num_real_particles,
**props)
def dump_output(self, *print_properties):
""" Print output based on level of detail required
The default detail level (low) is the integrator's calc's update
property for each named particle array.
The higher detail level dumps all particle array properties.
"""
fname = self.fname + "_"
props = {}
cell_size = self.particles.cell_manager.cell_size
for pa in self.particles.arrays:
name = pa.name
_fname = os.path.join(self.output_directory, fname + name + "_" + str(self.t) + ".npz")
if self.detailed_output:
savez(_fname, dt=self.dt, **pa.properties)
else:
for prop in print_properties:
props[prop] = pa.get(prop)
savez(_fname, dt=self.dt, cell_size=cell_size, np=pa.num_real_particles, **props)
def save_current_genealogy(self):
savefpath = os.path.join(self._logdir,
f'genealogy_gen{self._istep:03d}.npz')
save_kwargs = {
'image_ids': np.array(self._curr_sample_ids, dtype=str),
'genealogy': np.array(self._genealogy, dtype=str)
}
utils.savez(savefpath, save_kwargs)
ImageidTable[stepi][0:len(score_tmp)] = image_ids
except FileNotFoundError:
if stepi == 0:
startnum += 1
steps += 1
continue
else:
print("maximum steps is %d." % stepi)
ScoreEvolveTable = ScoreEvolveTable[0:stepi, :]
ImageidTable = ImageidTable[0:stepi]
steps = stepi
break
ImageidTable = np.asarray(ImageidTable)
utils.savez(os.path.join(CurDataDir, "scores_summary_table.npz"), {
"ScoreEvolveTable": ScoreEvolveTable,
"ImageidTable": ImageidTable
})
#%% Filter the Samples that has Score in a given range
def select_image(CurDataDir, lb=200, ub=None):
fncatalog = os.listdir(CurDataDir)
ScoreEvolveTable, ImageidTable = utils.scores_summary(CurDataDir)
# it will automatic read the existing summary or generate one.
if ub is None:
ub = np.nanmax(ScoreEvolveTable) + 1
if lb is None:
lb = np.nanmin(ScoreEvolveTable) - 1
imgid_list = ImageidTable[np.logical_and(ScoreEvolveTable > lb,
ScoreEvolveTable < ub)]
def dump_output(self, dt, *print_properties):
""" Print output based on level of detail required
The default detail level (low) is the integrator's calc's update
property for each named particle array.
The higher detail level dumps all particle array properties.
Format:
-------
A single file named as: <fname>_<rank>_<count>.npz
The output file contains the following fields:
solver_data : Solver related data like time step, time and
iteration count. These are used to resume a simulation.
arrays : A dictionary keyed on particle array names and with
particle properties as value.
version : The version number for this format of file
output. The current version number is 1
Example:
--------
data = load('foo.npz')
version = data['version']
dt = data['solver_data']['dt']
t = data['solver_data']['t']
array = data['arrays'][array_name].astype(object)
array['x']
"""
if self.with_cl:
self.particles.read_from_buffer()
fname = self.fname + '_'
props = {"arrays":{}, "solver_data":{}}
cell_size = None
if not self.with_cl:
cell_size = self.particles.cell_manager.cell_size
_fname = os.path.join(self.output_directory,
fname + str(self.count) +'.npz')
if self.detailed_output:
for array in self.particles.arrays:
props["arrays"][array.name]=array.get_property_arrays(all=True)
else:
for array in self.particles.arrays:
props["arrays"][array.name]=array.get_property_arrays(all=False)
# Add the solver data
props["solver_data"]["dt"] = dt
props["solver_data"]["t"] = self.t
props["solver_data"]["count"] = self.count
if self.parallel_output_mode == "collected" and self.in_parallel:
comm = self.comm
arrays = props["arrays"]
numarrays = len(arrays)
array_names = arrays.keys()
# gather the data from all processors
collected_data = comm.gather(arrays, root=0)
if self.rank == 0:
props["arrays"] = {}
size = comm.Get_size()
# concatenate the arrays
for array_name in array_names:
props["arrays"][array_name] = {}
_props = collected_data[0][array_name].keys()
for prop in _props:
prop_arr = numpy.concatenate( [collected_data[pid][array_name][prop] for pid in range(size)] )
props["arrays"][array_name][prop] = prop_arr
savez(_fname, version=1, **props)
else:
savez(_fname, version=1, **props)