def finished(self):
"""Write elapsed time to log file and move the log file to the data
directory"""
from mpi4py.MPI import COMM_WORLD as comm
from mpi4py.MPI import Wtime
# Wait until all processors are done
comm.barrier()
seconds = Wtime() - self.wt
if comm.rank == 0:
# import subprocess
from datetime import datetime
# Time at end of simulation
i = datetime.now()
endtime = i.strftime("%d/%m/%Y at %H:%M:%S")
f = open(self.data_folder + "psecas.log", "a")
f.write("\nCalculation ended on " + endtime + "\n")
m, s = divmod(seconds, 60)
h, m = divmod(m, 60)
d, h = divmod(h, 24)
msg = (
"Time elapsed was {} days {} hours {} minutes {:1.4} seconds"
)
f.write(msg.format(d, h, m, s))
f.close()
def main(argv=None):
args = process_command_line(argv)
# note that in MPI mode, lengths will be global, whereas data will
# be local (i.e. only this node's data).
lengths, data = load_trjs_or_features(args)
kwargs = {}
if args.cluster_iterations is not None:
kwargs['kmedoids_updates'] = int(args.cluster_iterations)
clustering = args.Clusterer(metric=args.cluster_distance,
n_clusters=args.cluster_number,
cluster_radius=args.cluster_radius,
mpi_mode=mpi_mode,
**kwargs)
clustering.fit(data)
logger.info("Clustered %s frames into %s clusters in %s seconds.",
sum(lengths), len(clustering.centers_), clustering.runtime_)
result = clustering.result_
if mpi_mode:
local_ctr_inds, local_dists, local_assigs = \
result.center_indices, result.distances, result.assignments
with timed("Reassembled dist and assign arrays in %.2f sec",
logging.info):
all_dists = mpi.ops.assemble_striped_ragged_array(
local_dists, lengths)
all_assigs = mpi.ops.assemble_striped_ragged_array(
local_assigs, lengths)
ctr_inds = mpi.ops.convert_local_indices(local_ctr_inds, lengths)
result = ClusterResult(center_indices=ctr_inds,
distances=all_dists,
assignments=all_assigs,
centers=result.centers)
result = result.partition(lengths)
if mpi_comm.Get_rank() == 0:
write_centers_indices(args.center_indices,
[(t, f * args.subsample)
for t, f in result.center_indices])
write_centers(result, args)
write_assignments_and_distances_with_reassign(result, args)
mpi_comm.barrier()
logger.info("Success! Data can be found in %s.",
os.path.dirname(args.distances))
return 0
def __init__(self, system, data_folder, experiment, steps, tag=""):
"""
Initialisation creates the output directory and saves the path to the
io object.
It also copies the experiment script to the data directory and
saves a dictionary with important information.
"""
from mpi4py.MPI import COMM_WORLD as comm
from mpi4py.MPI import Wtime
from numpy import arange
self.system = system
self.steps = steps
self.index_global = arange(steps)
self.index_local = self.index_global[comm.rank :: comm.size]
self.steps_local = len(self.index_local)
# Folder where data is stored
assert data_folder[-1] == "/", "data_folder should end with /"
self.data_folder = data_folder
if comm.rank == 0:
import subprocess
from datetime import datetime
from numpy import float64
# Create datafolder
subprocess.call("mkdir " + self.data_folder, shell=True)
# Copy the experiment to the data folder
# experiment = path.basename(experiment)
subprocess.call(
"cp " + experiment + " " + data_folder + experiment,
shell=True,
)
info = {"experiment": experiment}
# Save git commit number
try:
git_commit = subprocess.check_output(["git", "rev-parse", "HEAD"])
git_commit = git_commit.strip().decode("utf-8")
info.update({"git_commit": git_commit})
except:
pass
# Save the hostname
hostname = subprocess.check_output(["hostname"])
hostname = hostname.strip().decode("utf-8")
info.update({"hostname": hostname})
# Save the time at the start of simulation
i = datetime.now()
simulation_start = i.strftime("%d/%m/%Y at %H:%M:%S")
info.update({"simulation_start": simulation_start})
# Add tag which can be used to group simulations together
info.update({"tag": tag})
# Collect all variables in local namespace that are int, float or
# float64. These will in general be the values set by us.
# for key in local_vars.keys():
# if type(local_vars[key]) in (float, float64, int):
# info.update({key: local_vars[key]})
# Save the number of MPI processes used
info.update({"MPI": comm.size})
# Save the dictionary info to the file info.p
# pickle.dump(info, open(data_folder+'info.p', 'wb'))
# Start a log file
f = open(self.data_folder + "psecas.log", "w")
f.write("Log file for EVP run\n")
f.write(
"Solving {} evp problems on {} processors\n\n".format(
self.steps, comm.size
)
)
# Write start date and time
f.write("Calculation started on " + simulation_start + "\n\n")
# Write the contents of info.p for convenience
f.write("Contents of info.p is printed below \n")
for key in info.keys():
f.write(key + " = {} \n".format(info[key]))
f.write("\nContents of system is printed below \n")
for key in self.system.__dict__.keys():
if type(self.system.__dict__[key]) in (float, int, list):
f.write(key + " = {}\n".format(self.system.__dict__[key]))
grid = self.system.grid
f.write("\nUsing {} with\n".format(type(grid)))
for key in grid.__dict__.keys():
if type(grid.__dict__[key]) in (float, float64, int):
f.write(key + " = {} \n".format(grid.__dict__[key]))
f.write("\n\nEntering main calculation loop \n")
f.close()
# Used for computing total runtime
self.wt = Wtime()
# Wait until all processors are done
comm.barrier()
def _create_tempdir(request, mode=None):
"""
Adapted from DOLFIN's dolfin_utils/test/fixtures.py.
"""
# Get directory name of test_foo.py file
testfile = request.module.__file__
testfiledir = os.path.dirname(os.path.abspath(testfile))
# Construct name test_foo_tempdir from name test_foo.py
testfilename = os.path.basename(testfile)
if hasattr(request.config, "slaveinput"):
outputname = testfilename.replace(
".py",
"_tempdir_{}".format(request.config.slaveinput["slaveid"]))
else:
outputname = testfilename.replace(".py", "_tempdir")
# Get function name test_something from test_foo.py
function = request.function.__name__
if mode == "save":
function = function.replace("_save", "_io")
elif mode == "load":
function = function.replace("_load", "_io")
# Join all of these to make a unique path for this test function
basepath = os.path.join(testfiledir, outputname)
path = os.path.join(basepath, function)
# Add a sequence number to avoid collisions when tests are otherwise parameterized
if COMM_WORLD.rank == 0:
_create_tempdir._sequencenumber[path] += 1
sequencenumber = _create_tempdir._sequencenumber[path]
sequencenumber = COMM_WORLD.allreduce(sequencenumber, op=SUM)
else:
sequencenumber = COMM_WORLD.allreduce(0, op=SUM)
path += "__" + str(sequencenumber)
# Delete and re-create directory on root node
if COMM_WORLD.rank == 0:
# First time visiting this basepath, delete the old and create
# a new if mode is not load
if basepath not in _create_tempdir._basepaths:
_create_tempdir._basepaths.add(basepath)
if mode == "load":
assert os.path.exists(basepath)
else:
if os.path.exists(basepath):
shutil.rmtree(basepath)
# Make sure we have the base path test_foo_tempdir for
# this test_foo.py file
if not os.path.exists(basepath):
os.mkdir(basepath)
# Delete path from old test run if mode is not load
if mode == "load":
assert os.path.exists(path)
else:
if os.path.exists(path):
shutil.rmtree(path)
# Make sure we have the path for this test execution:
# e.g. test_foo_tempdir/test_something__3
if not os.path.exists(path):
os.mkdir(path)
COMM_WORLD.barrier()
return path
os.mkdir('./vtk/')
recs = os.listdir(path)
recs.sort()
if np.mod(len(recs) - 1, comm.size) != 0:
if comm.rank == 0:
print('Number of records is not dividable by number of procs ')
recs = recs[comm.rank::comm.size]
if selection == 'latest':
recs = [
recs[-1],
]
comm.barrier()
for rec in recs:
record = h5py.File(path + rec, 'r')
base_str = '/data'
flds_str = '/fields/'
parts_str = '/species/'
info_str = '/info/'
Args = {}
iter = int(rec.split('.')[0])
iter_str = str(iter)
for key in ['Xgrid', 'Rgrid', 'dx', 'M', 'dt', 'FrameVelocity']:
Args[key] = record[base_str + info_str + key].value