def test_save_and_load_model_pmi(self):
pmi = {('ъ', 'ь'): 0.42}
with tempfile.TemporaryDirectory() as temp_dir:
path = os.path.join(temp_dir, 'model')
save_model(path, 'pmi', pmi)
model = load_model(path)
self.assertEqual(len(model), 2)
self.assertEqual(model[0], 'pmi')
self.assertEqual(model[1], pmi)
def run(self, raw_args=None):
"""
Parse the given args (if these are None, default to parsing sys.argv,
which is what you would want unless you are unit testing).
"""
args = self.parser.parse_args(raw_args)
output_cldf_dataset = False
start_time = time.time()
if args.ipa:
sc = pyclts.SoundClasses(args.sound_class_model)
def transform(sound):
return sc[sound]
else:
transform = None
try:
try:
dataset = CLDFDataset(args.dataset, transform=transform)
output_cldf_dataset = True
except JSONDecodeError:
dataset = Dataset(args.dataset,
args.dialect_input,
transform=transform)
algorithm, model = load_model(args.model)
except (DatasetError, ModelError) as err:
self.parser.error(str(err))
print('running {} on {}, conversion={}{}{}'.format(
args.model, args.dataset, 'ipa→' if args.ipa else '(',
args.sound_class_model, '' if args.ipa else ')'))
if algorithm == 'phmm':
scores = apply_phmm(dataset, *model)
else:
scores = apply_pmi(dataset, model)
clusters = cluster(dataset, scores, method=args.cluster_method)
if output_cldf_dataset:
write_cldf_clusters(clusters, args.output, dataset.dataset)
else:
write_clusters(clusters, args.output, args.dialect_output)
if args.time:
print('running time: {:.2f} sec'.format(time.time() - start_time))
if args.evaluate:
score = calc_f_score(dataset.get_clusters(), clusters)
print('f-score: {:.4f}'.format(score))
def test_save_and_load_model_phmm(self):
em = np.array([0.1, 0.2])
gx = np.array([0.3, 0.4])
gy = np.array([0.5, 0.6])
trans = np.array([0.7, 0.8])
with tempfile.TemporaryDirectory() as temp_dir:
path = os.path.join(temp_dir, 'model')
save_model(path, 'phmm', [em, gx, gy, trans])
model = load_model(path)
self.assertEqual(len(model), 2)
self.assertEqual(model[0], 'phmm')
for index, param in enumerate([em, gx, gy, trans]):
self.assertTrue(type(model[1][index]) is np.ndarray)
np.testing.assert_array_equal(model[1][index], param)
def test_load_model_with_bad_file(self):
with self.assertRaises(ModelError) as cm:
load_model(os.path.abspath(__file__))
self.assertTrue(str(cm.exception).startswith('Could not read model'))
def test_load_model_with_bad_path(self):
with self.assertRaises(ModelError) as cm:
load_model('')
self.assertTrue(str(cm.exception).startswith('Could not open model'))