rewrite tests with pytest

.github/workflows/ci.yaml

@@ -58,4 +58,4 @@ jobs:
         pysen run lint
     - name: Test
       run: |
-        nosetests --with-coverage --cover-package=nnmnkwii -v -w tests/ -a '!require_local_data,!modspec'
+        pytest --cov=nnmnkwii --cov-report xml -v tests/

setup.py

@@ -159,10 +159,10 @@ def package_files(directory):
     ext_modules=ext_modules,
     cmdclass=cmdclass,
     install_requires=install_requires,
-    tests_require=["nose", "coverage"],
+    tests_require=["pytest", "coverage"],
     extras_require={
         "docs": ["numpydoc", "sphinx_rtd_theme"],
-        "test": ["nose", "pyworld", "librosa"],
+        "test": ["pytest", "pyworld", "librosa"],
         "lint": [
             "pysen",
             "types-setuptools",

tests/test_autograd.py

@@ -3,8 +3,6 @@
 from nnmnkwii import autograd as AF
 from nnmnkwii import paramgen as G
 from nnmnkwii.autograd._impl.mlpg import MLPG, UnitVarianceMLPG
-from nnmnkwii.autograd._impl.modspec import ModSpec
-from nose.plugins.attrib import attr
 from torch import nn
 from torch.autograd import gradcheck
 
@@ -220,23 +218,21 @@ def test_mlpg_variance_expand():
         assert np.allclose(y.data.numpy(), y_hat.data.numpy())
 
 
-@attr("modspec")
-def test_modspec_gradcheck():
-    static_dim = 12
-    T = 16
-    torch.manual_seed(1234)
-    n = 16
-    for norm in [None, "ortho"]:
-        inputs = (torch.rand(T, static_dim, requires_grad=True), n, norm)
-        assert gradcheck(ModSpec.apply, inputs, eps=1e-4, atol=1e-4)
-
-
-@attr("modspec")
-def test_modspec_gradcheck_large_n():
-    static_dim = 12
-    T = 16
-    torch.manual_seed(1234)
-    for n in [16, 32]:
-        for norm in [None, "ortho"]:
-            inputs = (torch.rand(T, static_dim, requires_grad=True), n, norm)
-            assert gradcheck(ModSpec.apply, inputs, eps=1e-4, atol=1e-4)
+# def test_modspec_gradcheck():
+#     static_dim = 12
+#     T = 16
+#     torch.manual_seed(1234)
+#     n = 16
+#     for norm in [None, "ortho"]:
+#         inputs = (torch.rand(T, static_dim, requires_grad=True), n, norm)
+#         assert gradcheck(ModSpec.apply, inputs, eps=1e-4, atol=1e-4)
+
+
+# def test_modspec_gradcheck_large_n():
+#     static_dim = 12
+#     T = 16
+#     torch.manual_seed(1234)
+#     for n in [16, 32]:
+#         for norm in [None, "ortho"]:
+#             inputs = (torch.rand(T, static_dim, requires_grad=True), n, norm)
+#             assert gradcheck(ModSpec.apply, inputs, eps=1e-4, atol=1e-4)

tests/test_baseline.py

@@ -1,7 +1,6 @@
 from os.path import dirname, join
 
 import numpy as np
-from nose.plugins.attrib import attr
 from numpy.linalg import norm
 from sklearn.mixture import GaussianMixture
 
@@ -29,7 +28,6 @@ def _get_windows_set():
     return windows_set
 
 
-@attr("requires_bandmat")
 def test_diffvc():
     from nnmnkwii.baseline.gmm import MLPG
 
@@ -62,7 +60,6 @@ def test_diffvc():
         assert norm(tgt_mc - mc_converted1) < norm(src_mc - mc_converted1)
 
 
-@attr("requires_bandmat")
 def test_gmmmap_swap():
     from nnmnkwii.baseline.gmm import MLPG
 

tests/test_datasets.py

@@ -1,6 +1,7 @@
 from os.path import dirname, join
 
 import numpy as np
+import pytest
 from nnmnkwii.datasets import (
     FileDataSource,
     FileSourceDataset,
@@ -11,8 +12,6 @@
     example_file_data_sources_for_acoustic_model,
     example_file_data_sources_for_duration_model,
 )
-from nose.plugins.attrib import attr
-from nose.tools import raises
 
 DATA_DIR = join(dirname(__file__), "data")
 
@@ -45,7 +44,8 @@ def __test_outof_range(X):
         print(X[0])
 
     # Should raise IndexError
-    yield raises(IndexError)(__test_outof_range), X
+    with pytest.raises(IndexError):
+        __test_outof_range(X)
 
 
 def test_invalid_dataset():
@@ -71,19 +71,19 @@ def __test_wrong_num_collected_files():
         X = FileSourceDataset(WrongNumberOfCollectedFilesDataSource())
         X[0]
 
-    yield raises(TypeError)(__test_wrong_num_args)
-    yield raises(ValueError)(__test_wrong_num_collected_files)
+    with pytest.raises(TypeError):
+        __test_wrong_num_args()
+    with pytest.raises(ValueError):
+        __test_wrong_num_collected_files()
 
 
-@attr("pickle")
 def test_asarray_tqdm():
     # verbose=1 triggers tqdm progress report
     for padded in [True, False]:
         X, _ = _get_small_datasets(padded=padded, duration=True)
         X.asarray(verbose=1)
 
 
-@attr("pickle")
 def test_asarray():
     X, Y = _get_small_datasets(padded=False, duration=True)
     lengths = [len(x) for x in X]
@@ -110,17 +110,16 @@ def __test_very_small_padded_length():
         X.asarray(padded_length=1)
 
     # Should raise `num frames exceeded`
-    yield raises(RuntimeError)(__test_very_small_padded_length)
+    with pytest.raises(RuntimeError):
+        __test_very_small_padded_length()
 
 
-@attr("pickle")
 def test_duration_sources():
     X, Y = _get_small_datasets(padded=False, duration=True)
     for idx, (x, y) in enumerate(zip(X, Y)):
         print(idx, x.shape, y.shape)
 
 
-@attr("pickle")
 def test_slice():
     X, _ = _get_small_datasets(padded=False)
     x = X[:2]
@@ -133,14 +132,12 @@ def test_slice():
     assert len(x.shape) == 3 and x.shape[0] == 2
 
 
-@attr("pickle")
 def test_variable_length_sequence_wise_iteration():
     X, Y = _get_small_datasets(padded=False)
     for idx, (x, y) in enumerate(zip(X, Y)):
         print(idx, x.shape, y.shape)
 
 
-@attr("pickle")
 def test_fixed_length_sequence_wise_iteration():
     X, Y = _get_small_datasets(padded=True)
 
@@ -153,7 +150,6 @@ def test_fixed_length_sequence_wise_iteration():
         assert y.shape[0] == Ty
 
 
-@attr("pickle")
 def test_frame_wise_iteration():
     X, Y = _get_small_datasets(padded=False)
 
@@ -181,7 +177,6 @@ def test_frame_wise_iteration():
         pass
 
 
-@attr("pickle")
 def test_sequence_wise_torch_data_loader():
     import torch
     from torch.utils import data as data_utils
@@ -210,19 +205,20 @@ def __test(X, Y, batch_size):
             print(idx, x.shape, y.shape)
 
     # Test with batch_size = 1
-    yield __test, X, Y, 1
+    __test(X, Y, 1)
     # Since we have variable length frames, batch size larger than 1 causes
     # runtime error.
-    yield raises(RuntimeError)(__test), X, Y, 2
+    with pytest.raises(RuntimeError):
+        __test(X, Y, 2)
 
     # For padded dataset, which can be reprensented by (N, T^max, D), batchsize
     # can be any number.
     X, Y = _get_small_datasets(padded=True)
-    yield __test, X, Y, 1
-    yield __test, X, Y, 2
+    __test(X, Y, 1)
+    __test(X, Y, 2)
 
 
-@attr("pickle")
+# @attr("pickle")
 def test_frame_wise_torch_data_loader():
     import torch
     from torch.utils import data as data_utils
@@ -259,5 +255,5 @@ def __test(X, Y, batch_size):
             assert len(x.shape) == 2
             assert len(y.shape) == 2
 
-    yield __test, X, Y, 128
-    yield __test, X, Y, 256
+    __test(X, Y, 128)
+    __test(X, Y, 256)

tests/test_frontend.py

@@ -1,10 +1,10 @@
 from os.path import dirname, join
 
 import numpy as np
+import pytest
 from nnmnkwii.frontend import merlin as fe
 from nnmnkwii.io import hts
 from nnmnkwii.util import example_label_file, example_question_file
-from nose.tools import raises
 
 DATA_DIR = join(dirname(__file__), "data")
 
@@ -14,7 +14,7 @@ def test_invalid_linguistic_features():
     phone_labels = hts.load(example_label_file(phone_level=True))
     state_labels = hts.load(example_label_file(phone_level=False))
 
-    @raises(ValueError)
+    # @raises(ValueError)
     def __test(labels, subphone_features, add_frame_features):
         fe.linguistic_features(
             labels,
@@ -24,19 +24,22 @@ def __test(labels, subphone_features, add_frame_features):
             add_frame_features=add_frame_features,
         )
 
-    yield __test, phone_labels, "full", True
-    yield __test, phone_labels, "full", False
-    yield __test, state_labels, "full", False
+    with pytest.raises(ValueError):
+        __test(phone_labels, "full", True)
+    with pytest.raises(ValueError):
+        __test(phone_labels, "full", False)
+    with pytest.raises(ValueError):
+        __test(state_labels, "full", False)
 
 
 def test_invalid_duration_features():
     phone_labels = hts.load(example_label_file(phone_level=True))
 
-    @raises(ValueError)
     def __test(labels, unit_size, feature_size):
         fe.duration_features(labels, unit_size=unit_size, feature_size=feature_size)
 
-    yield __test, phone_labels, None, "frame"
+    with pytest.raises(ValueError):
+        __test(phone_labels, None, "frame")
 
 
 def test_silence_frame_removal_given_hts_labels():

tests/test_io.py

@@ -2,10 +2,10 @@
 import re
 from os.path import dirname, join
 
+import pytest
 from nnmnkwii.frontend import merlin as fe
 from nnmnkwii.io import hts
 from nnmnkwii.util import example_question_file
-from nose.tools import raises
 
 try:
     import pyopenjtalk  # noqa
@@ -183,7 +183,6 @@ def test_hts_append():
         labels.append(label)
     assert str(test_labels) == str(labels)
 
-    @raises(ValueError)
     def test_invalid_start_time():
         labels = hts.HTSLabelFile()
         labels.append((100000, 0, "NG"))
@@ -193,7 +192,6 @@ def test_succeeding_times():
         labels.append((0, 1000000, "OK"))
         labels.append((1000000, 2000000, "OK"))
 
-    @raises(ValueError)
     def test_non_succeeding_times():
         labels = hts.HTSLabelFile()
         labels.append((0, 1000000, "OK"))
@@ -204,9 +202,11 @@ def test_non_succeeding_times_wo_strict():
         labels.append((0, 1000000, "OK"), strict=False)
         labels.append((1500000, 2000000, "OK"), strict=False)
 
-    test_invalid_start_time()
+    with pytest.raises(ValueError):
+        test_invalid_start_time()
     test_succeeding_times()
-    test_non_succeeding_times()
+    with pytest.raises(ValueError):
+        test_non_succeeding_times()
     test_non_succeeding_times_wo_strict()
 
 
@@ -227,20 +227,20 @@ def test_create_from_contexts():
     labels2 = hts.HTSLabelFile.create_from_contexts(contexts)
     assert str(labels), str(labels2)
 
-    @raises(ValueError)
     def test_empty_context():
         hts.HTSLabelFile.create_from_contexts("")
 
-    @raises(ValueError)
     def test_empty_context2():
         contexts = pyopenjtalk.extract_fullcontext("")
         hts.HTSLabelFile.create_from_contexts(contexts)
 
-    test_empty_context()
+    with pytest.raises(ValueError):
+        test_empty_context()
     try:
         import pyopenjtalk  # noqa
 
-        test_empty_context2()
+        with pytest.raises(ValueError):
+            test_empty_context2()
     except ImportError:
         pass
 

tests/test_metrics.py

@@ -28,8 +28,8 @@ def __test(f):
             np.testing.assert_almost_equal(f(x, y, lengths), f(x, y), decimal=5)
             assert f(x, y) > 0
 
-    yield __test, metrics.melcd
-    yield __test, metrics.mean_squared_error
+    __test(metrics.melcd)
+    __test(metrics.mean_squared_error)
 
 
 def test_f0_mse():

tests/test_paramgen.py

@@ -1,5 +1,4 @@
 import numpy as np
-from nose.plugins.attrib import attr
 
 
 def _get_windows_set():
@@ -29,7 +28,6 @@ def _get_windows_set():
     return windows_set
 
 
-@attr("requires_bandmat")
 def test_mlpg():
     from nnmnkwii import paramgen as G
 
@@ -61,7 +59,6 @@ def test_mlpg():
             assert np.allclose(generated1, generated2)
 
 
-@attr("requires_bandmat")
 def test_mlpg_window_full():
     from nnmnkwii import paramgen as G
 
@@ -82,7 +79,6 @@ def full_window_mat_native(win_mats, T):
         assert np.allclose(full_window_mat_native(win_mats, T), fullwin)
 
 
-@attr("requires_bandmat")
 def test_unit_variance_mlpg():
     from nnmnkwii import paramgen as G
 
@@ -99,7 +95,6 @@ def test_unit_variance_mlpg():
         assert np.allclose(y_hat, y)
 
 
-@attr("requires_bandmat")
 def test_reshape_means():
     from nnmnkwii import paramgen as G