def test_1(self):
vtraj = self.vtrajs[2]
assert len(vtraj) == 3
assignments = 1234 * np.ones(len(vtraj))
distances = np.random.randn(len(vtraj)).astype(np.float32)
save(self.fa, self.fd, assignments, distances, vtraj)
AData = -1 * np.ones((2,10))
AData[0,6:9] = 1234
DData = -1 * np.ones((2,10), dtype=np.float32)
DData[0,6:9] = distances
npt.assert_equal(self.fa.root.Data, AData)
npt.assert_equal(self.fd.root.Data, DData)
def main(args, logger):
metric = construct_metric(args)
project = Project.LoadFromHDF(args.project)
if not os.path.exists(args.generators):
raise IOError('Could not open generators')
generators = os.path.abspath(args.generators)
output_dir = os.path.abspath(args.output_dir)
# connect to the workers
try:
json_file = client_json_file(args.profile, args.cluster_id)
client = parallel.Client(json_file, timeout=2)
except parallel.error.TimeoutError as exception:
msg = '\nparallel.error.TimeoutError: ' + str(exception)
msg += "\n\nPerhaps you didn't start a controller?\n"
msg += "(hint, use ipcluster start)"
print >> sys.stderr, msg
sys.exit(1)
lview = client.load_balanced_view()
# partition the frames into a bunch of vtrajs
all_vtrajs = local.partition(project, args.chunk_size)
# initialze the containers to save to disk
f_assignments, f_distances = local.setup_containers(output_dir,
project, all_vtrajs)
# get the chunks that have not been computed yet
valid_indices = np.where(f_assignments.root.completed_vtrajs[:] == False)[0]
remaining_vtrajs = np.array(all_vtrajs)[valid_indices].tolist()
logger.info('%d/%d jobs remaining', len(remaining_vtrajs), len(all_vtrajs))
# send the workers the files they need to get started
# dview.apply_sync(remote.load_gens, generators, project['ConfFilename'],
# metric)
# get the workers going
n_jobs = len(remaining_vtrajs)
amr = lview.map(remote.assign, remaining_vtrajs,
[generators]*n_jobs, [metric]*n_jobs, chunksize=1)
pending = set(amr.msg_ids)
while pending:
client.wait(pending, 1e-3)
# finished is the set of msg_ids that are complete
finished = pending.difference(client.outstanding)
# update pending to exclude those that just finished
pending = pending.difference(finished)
for msg_id in finished:
# we know these are done, so don't worry about blocking
async = client.get_result(msg_id)
assignments, distances, chunk = async.result[0]
vtraj_id = local.save(f_assignments, f_distances, assignments, distances, chunk)
log_status(logger, len(pending), n_jobs, vtraj_id, async)
f_assignments.close()
f_distances.close()
logger.info('All done, exiting.')