1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
|
import io
from unittest.mock import patch
from torchaudio._internal import module_utils as _mod_utils
from torchaudio.backend import soundfile_backend
from torchaudio_unittest.common_utils import (
get_wav_data,
load_wav,
nested_params,
PytorchTestCase,
skipIfNoModule,
TempDirMixin,
)
from .common import fetch_wav_subtype, parameterize, skipIfFormatNotSupported
if _mod_utils.is_module_available("soundfile"):
import soundfile
class MockedSaveTest(PytorchTestCase):
@nested_params(
["float32", "int32", "int16", "uint8"],
[8000, 16000],
[1, 2],
[False, True],
[
(None, None),
("PCM_U", None),
("PCM_U", 8),
("PCM_S", None),
("PCM_S", 16),
("PCM_S", 32),
("PCM_F", None),
("PCM_F", 32),
("PCM_F", 64),
("ULAW", None),
("ULAW", 8),
("ALAW", None),
("ALAW", 8),
],
)
@patch("soundfile.write")
def test_wav(self, dtype, sample_rate, num_channels, channels_first, enc_params, mocked_write):
"""soundfile_backend.save passes correct subtype to soundfile.write when WAV"""
filepath = "foo.wav"
input_tensor = get_wav_data(
dtype,
num_channels,
num_frames=3 * sample_rate,
normalize=dtype == "float32",
channels_first=channels_first,
).t()
encoding, bits_per_sample = enc_params
soundfile_backend.save(
filepath,
input_tensor,
sample_rate,
channels_first=channels_first,
encoding=encoding,
bits_per_sample=bits_per_sample,
)
# on +Py3.8 call_args.kwargs is more descreptive
args = mocked_write.call_args[1]
assert args["file"] == filepath
assert args["samplerate"] == sample_rate
assert args["subtype"] == fetch_wav_subtype(dtype, encoding, bits_per_sample)
assert args["format"] is None
self.assertEqual(args["data"], input_tensor.t() if channels_first else input_tensor)
@patch("soundfile.write")
def assert_non_wav(
self,
fmt,
dtype,
sample_rate,
num_channels,
channels_first,
mocked_write,
encoding=None,
bits_per_sample=None,
):
"""soundfile_backend.save passes correct subtype and format to soundfile.write when SPHERE"""
filepath = f"foo.{fmt}"
input_tensor = get_wav_data(
dtype,
num_channels,
num_frames=3 * sample_rate,
normalize=False,
channels_first=channels_first,
).t()
expected_data = input_tensor.t() if channels_first else input_tensor
soundfile_backend.save(
filepath,
input_tensor,
sample_rate,
channels_first,
encoding=encoding,
bits_per_sample=bits_per_sample,
)
# on +Py3.8 call_args.kwargs is more descreptive
args = mocked_write.call_args[1]
assert args["file"] == filepath
assert args["samplerate"] == sample_rate
if fmt in ["sph", "nist", "nis"]:
assert args["format"] == "NIST"
else:
assert args["format"] is None
self.assertEqual(args["data"], expected_data)
@nested_params(
["sph", "nist", "nis"],
["int32", "int16"],
[8000, 16000],
[1, 2],
[False, True],
[
("PCM_S", 8),
("PCM_S", 16),
("PCM_S", 24),
("PCM_S", 32),
("ULAW", 8),
("ALAW", 8),
("ALAW", 16),
("ALAW", 24),
("ALAW", 32),
],
)
def test_sph(self, fmt, dtype, sample_rate, num_channels, channels_first, enc_params):
"""soundfile_backend.save passes default format and subtype (None-s) to
soundfile.write when not WAV"""
encoding, bits_per_sample = enc_params
self.assert_non_wav(
fmt, dtype, sample_rate, num_channels, channels_first, encoding=encoding, bits_per_sample=bits_per_sample
)
@parameterize(
["int32", "int16"],
[8000, 16000],
[1, 2],
[False, True],
[8, 16, 24],
)
def test_flac(self, dtype, sample_rate, num_channels, channels_first, bits_per_sample):
"""soundfile_backend.save passes default format and subtype (None-s) to
soundfile.write when not WAV"""
self.assert_non_wav("flac", dtype, sample_rate, num_channels, channels_first, bits_per_sample=bits_per_sample)
@parameterize(
["int32", "int16"],
[8000, 16000],
[1, 2],
[False, True],
)
def test_ogg(self, dtype, sample_rate, num_channels, channels_first):
"""soundfile_backend.save passes default format and subtype (None-s) to
soundfile.write when not WAV"""
self.assert_non_wav("ogg", dtype, sample_rate, num_channels, channels_first)
@skipIfNoModule("soundfile")
class SaveTestBase(TempDirMixin, PytorchTestCase):
def assert_wav(self, dtype, sample_rate, num_channels, num_frames):
"""`soundfile_backend.save` can save wav format."""
path = self.get_temp_path("data.wav")
expected = get_wav_data(dtype, num_channels, num_frames=num_frames, normalize=False)
soundfile_backend.save(path, expected, sample_rate)
found, sr = load_wav(path, normalize=False)
assert sample_rate == sr
self.assertEqual(found, expected)
def _assert_non_wav(self, fmt, dtype, sample_rate, num_channels):
"""`soundfile_backend.save` can save non-wav format.
Due to precision missmatch, and the lack of alternative way to decode the
resulting files without using soundfile, only meta data are validated.
"""
num_frames = sample_rate * 3
path = self.get_temp_path(f"data.{fmt}")
expected = get_wav_data(dtype, num_channels, num_frames=num_frames, normalize=False)
soundfile_backend.save(path, expected, sample_rate)
sinfo = soundfile.info(path)
assert sinfo.format == fmt.upper()
assert sinfo.frames == num_frames
assert sinfo.channels == num_channels
assert sinfo.samplerate == sample_rate
def assert_flac(self, dtype, sample_rate, num_channels):
"""`soundfile_backend.save` can save flac format."""
self._assert_non_wav("flac", dtype, sample_rate, num_channels)
def assert_sphere(self, dtype, sample_rate, num_channels):
"""`soundfile_backend.save` can save sph format."""
self._assert_non_wav("nist", dtype, sample_rate, num_channels)
def assert_ogg(self, dtype, sample_rate, num_channels):
"""`soundfile_backend.save` can save ogg format.
As we cannot inspect the OGG format (it's lossy), we only check the metadata.
"""
self._assert_non_wav("ogg", dtype, sample_rate, num_channels)
@skipIfNoModule("soundfile")
class TestSave(SaveTestBase):
@parameterize(
["float32", "int32", "int16"],
[8000, 16000],
[1, 2],
)
def test_wav(self, dtype, sample_rate, num_channels):
"""`soundfile_backend.save` can save wav format."""
self.assert_wav(dtype, sample_rate, num_channels, num_frames=None)
@parameterize(
["float32", "int32", "int16"],
[4, 8, 16, 32],
)
def test_multiple_channels(self, dtype, num_channels):
"""`soundfile_backend.save` can save wav with more than 2 channels."""
sample_rate = 8000
self.assert_wav(dtype, sample_rate, num_channels, num_frames=None)
@parameterize(
["int32", "int16"],
[8000, 16000],
[1, 2],
)
@skipIfFormatNotSupported("NIST")
def test_sphere(self, dtype, sample_rate, num_channels):
"""`soundfile_backend.save` can save sph format."""
self.assert_sphere(dtype, sample_rate, num_channels)
@parameterize(
[8000, 16000],
[1, 2],
)
@skipIfFormatNotSupported("FLAC")
def test_flac(self, sample_rate, num_channels):
"""`soundfile_backend.save` can save flac format."""
self.assert_flac("float32", sample_rate, num_channels)
@parameterize(
[8000, 16000],
[1, 2],
)
@skipIfFormatNotSupported("OGG")
def test_ogg(self, sample_rate, num_channels):
"""`soundfile_backend.save` can save ogg/vorbis format."""
self.assert_ogg("float32", sample_rate, num_channels)
@skipIfNoModule("soundfile")
class TestSaveParams(TempDirMixin, PytorchTestCase):
"""Test the correctness of optional parameters of `soundfile_backend.save`"""
@parameterize([True, False])
def test_channels_first(self, channels_first):
"""channels_first swaps axes"""
path = self.get_temp_path("data.wav")
data = get_wav_data("int32", 2, channels_first=channels_first)
soundfile_backend.save(path, data, 8000, channels_first=channels_first)
found = load_wav(path)[0]
expected = data if channels_first else data.transpose(1, 0)
self.assertEqual(found, expected, atol=1e-4, rtol=1e-8)
@skipIfNoModule("soundfile")
class TestFileObject(TempDirMixin, PytorchTestCase):
def _test_fileobj(self, ext):
"""Saving audio to file-like object works"""
sample_rate = 16000
path = self.get_temp_path(f"test.{ext}")
subtype = "FLOAT" if ext == "wav" else None
data = get_wav_data("float32", num_channels=2)
soundfile.write(path, data.numpy().T, sample_rate, subtype=subtype)
expected = soundfile.read(path, dtype="float32")[0]
fileobj = io.BytesIO()
soundfile_backend.save(fileobj, data, sample_rate, format=ext)
fileobj.seek(0)
found, sr = soundfile.read(fileobj, dtype="float32")
assert sr == sample_rate
self.assertEqual(expected, found, atol=1e-4, rtol=1e-8)
def test_fileobj_wav(self):
"""Saving audio via file-like object works"""
self._test_fileobj("wav")
@skipIfFormatNotSupported("FLAC")
def test_fileobj_flac(self):
"""Saving audio via file-like object works"""
self._test_fileobj("flac")
@skipIfFormatNotSupported("NIST")
def test_fileobj_nist(self):
"""Saving audio via file-like object works"""
self._test_fileobj("NIST")
@skipIfFormatNotSupported("OGG")
def test_fileobj_ogg(self):
"""Saving audio via file-like object works"""
self._test_fileobj("OGG")
|
