test.py 7.07 KB
Newer Older
1
from __future__ import absolute_import, division, print_function, unicode_literals
David Pollack's avatar
David Pollack committed
2
import unittest
3
import common_utils
4
import torch
Soumith Chintala's avatar
Soumith Chintala committed
5
import torchaudio
David Pollack's avatar
David Pollack committed
6
import math
David Pollack's avatar
David Pollack committed
7
8
import os

Soumith Chintala's avatar
Soumith Chintala committed
9

David Pollack's avatar
David Pollack committed
10
class Test_LoadSave(unittest.TestCase):
11
    test_dirpath, test_dir = common_utils.create_temp_assets_dir()
12
13
    test_filepath = os.path.join(test_dirpath, "assets",
                                 "steam-train-whistle-daniel_simon.mp3")
Soumith Chintala's avatar
Soumith Chintala committed
14

David Pollack's avatar
David Pollack committed
15
    def test_1_save(self):
David Pollack's avatar
David Pollack committed
16
        # load signal
David Pollack's avatar
David Pollack committed
17
        x, sr = torchaudio.load(self.test_filepath, normalization=False)
David Pollack's avatar
David Pollack committed
18
19
20
21
22
23
24
25
26
27
28
29
30
31

        # check save
        new_filepath = os.path.join(self.test_dirpath, "test.wav")
        torchaudio.save(new_filepath, x, sr)
        self.assertTrue(os.path.isfile(new_filepath))
        os.unlink(new_filepath)

        # check automatic normalization
        x /= 1 << 31
        torchaudio.save(new_filepath, x, sr)
        self.assertTrue(os.path.isfile(new_filepath))
        os.unlink(new_filepath)

        # test save 1d tensor
David Pollack's avatar
David Pollack committed
32
        x = x[0, :]  # get mono signal
Soumith Chintala's avatar
Soumith Chintala committed
33
        x.squeeze_()  # remove channel dim
David Pollack's avatar
David Pollack committed
34
35
36
37
38
39
        torchaudio.save(new_filepath, x, sr)
        self.assertTrue(os.path.isfile(new_filepath))
        os.unlink(new_filepath)

        # don't allow invalid sizes as inputs
        with self.assertRaises(ValueError):
David Pollack's avatar
David Pollack committed
40
            x.unsqueeze_(1)  # L x C not C x L
David Pollack's avatar
David Pollack committed
41
42
43
44
45
            torchaudio.save(new_filepath, x, sr)

        with self.assertRaises(ValueError):
            x.squeeze_()
            x.unsqueeze_(1)
Soumith Chintala's avatar
Soumith Chintala committed
46
            x.unsqueeze_(0)  # 1 x L x 1
David Pollack's avatar
David Pollack committed
47
48
49
50
            torchaudio.save(new_filepath, x, sr)

        # don't save to folders that don't exist
        with self.assertRaises(OSError):
51
52
            new_filepath = os.path.join(self.test_dirpath, "no-path",
                                        "test.wav")
David Pollack's avatar
David Pollack committed
53
            torchaudio.save(new_filepath, x, sr)
Soumith Chintala's avatar
Soumith Chintala committed
54

55
        # save created file
56
57
        sinewave_filepath = os.path.join(self.test_dirpath, "assets",
                                         "sinewave.wav")
58
59
60
61
        sr = 16000
        freq = 440
        volume = 0.3

62
        y = (torch.cos(
63
            2 * math.pi * torch.arange(0, 4 * sr).float() * freq / sr))
David Pollack's avatar
David Pollack committed
64
        y.unsqueeze_(0)
65
        # y is between -1 and 1, so must scale
David Pollack's avatar
David Pollack committed
66
        y = (y * volume * (2**31)).long()
67
68
        torchaudio.save(sinewave_filepath, y, sr)
        self.assertTrue(os.path.isfile(sinewave_filepath))
69

70
        # test precision
David Pollack's avatar
David Pollack committed
71
        new_precision = 32
72
        new_filepath = os.path.join(self.test_dirpath, "test.wav")
David Pollack's avatar
David Pollack committed
73
74
75
76
77
        si, ei = torchaudio.info(sinewave_filepath)
        torchaudio.save(new_filepath, y, sr, new_precision)
        si32, ei32 = torchaudio.info(new_filepath)
        self.assertEqual(si.precision, 16)
        self.assertEqual(si32.precision, new_precision)
78
79
        os.unlink(new_filepath)

David Pollack's avatar
David Pollack committed
80
81
82
83
84
85
    def test_2_load(self):
        # check normal loading
        x, sr = torchaudio.load(self.test_filepath)
        self.assertEqual(sr, 44100)
        self.assertEqual(x.size(), (2, 278756))

David Pollack's avatar
David Pollack committed
86
87
88
89
90
91
92
93
94
        # check no normalizing
        x, _ = torchaudio.load(self.test_filepath, normalization=False)
        self.assertTrue(x.min() <= -1.0)
        self.assertTrue(x.max() >= 1.0)

        # check offset
        offset = 15
        x, _ = torchaudio.load(self.test_filepath)
        x_offset, _ = torchaudio.load(self.test_filepath, offset=offset)
95
        self.assertTrue(x[:, offset:].allclose(x_offset))
David Pollack's avatar
David Pollack committed
96
97
98
99
100
101
102
103
104
105
106
107
108

        # check number of frames
        n = 201
        x, _ = torchaudio.load(self.test_filepath, num_frames=n)
        self.assertTrue(x.size(), (2, n))

        # check channels first
        x, _ = torchaudio.load(self.test_filepath, channels_first=False)
        self.assertEqual(x.size(), (278756, 2))

        # check different input tensor type
        x, _ = torchaudio.load(self.test_filepath, torch.LongTensor(), normalization=False)
        self.assertTrue(isinstance(x, torch.LongTensor))
David Pollack's avatar
David Pollack committed
109
110
111
112
113
114
115
116
117
118
119

        # check raising errors
        with self.assertRaises(OSError):
            torchaudio.load("file-does-not-exist.mp3")

        with self.assertRaises(OSError):
            tdir = os.path.join(
                os.path.dirname(self.test_dirpath), "torchaudio")
            torchaudio.load(tdir)

    def test_3_load_and_save_is_identity(self):
120
121
122
123
124
125
126
127
128
        input_path = os.path.join(self.test_dirpath, 'assets', 'sinewave.wav')
        tensor, sample_rate = torchaudio.load(input_path)
        output_path = os.path.join(self.test_dirpath, 'test.wav')
        torchaudio.save(output_path, tensor, sample_rate)
        tensor2, sample_rate2 = torchaudio.load(output_path)
        self.assertTrue(tensor.allclose(tensor2))
        self.assertEqual(sample_rate, sample_rate2)
        os.unlink(output_path)

David Pollack's avatar
David Pollack committed
129
    def test_4_load_partial(self):
David Pollack's avatar
David Pollack committed
130
131
        num_frames = 101
        offset = 201
132
133
134
135
        # load entire mono sinewave wav file, load a partial copy and then compare
        input_sine_path = os.path.join(self.test_dirpath, 'assets', 'sinewave.wav')
        x_sine_full, sr_sine = torchaudio.load(input_sine_path)
        x_sine_part, _ = torchaudio.load(input_sine_path, num_frames=num_frames, offset=offset)
136
        l1_error = x_sine_full[:, offset:(num_frames + offset)].sub(x_sine_part).abs().sum().item()
137
        # test for the correct number of samples and that the correct portion was loaded
David Pollack's avatar
David Pollack committed
138
        self.assertEqual(x_sine_part.size(1), num_frames)
139
140
141
142
143
144
145
        self.assertEqual(l1_error, 0.)
        # create a two channel version of this wavefile
        x_2ch_sine = x_sine_full.repeat(1, 2)
        out_2ch_sine_path = os.path.join(self.test_dirpath, 'assets', '2ch_sinewave.wav')
        torchaudio.save(out_2ch_sine_path, x_2ch_sine, sr_sine)
        x_2ch_sine_load, _ = torchaudio.load(out_2ch_sine_path, num_frames=num_frames, offset=offset)
        os.unlink(out_2ch_sine_path)
David Pollack's avatar
David Pollack committed
146
        l1_error = x_2ch_sine_load.sub(x_2ch_sine[:, offset:(offset + num_frames)]).abs().sum().item()
147
148
149
150
151
        self.assertEqual(l1_error, 0.)

        # test with two channel mp3
        x_2ch_full, sr_2ch = torchaudio.load(self.test_filepath, normalization=True)
        x_2ch_part, _ = torchaudio.load(self.test_filepath, normalization=True, num_frames=num_frames, offset=offset)
152
        l1_error = x_2ch_full[:, offset:(offset + num_frames)].sub(x_2ch_part).abs().sum().item()
David Pollack's avatar
David Pollack committed
153
        self.assertEqual(x_2ch_part.size(1), num_frames)
154
155
156
157
158
        self.assertEqual(l1_error, 0.)

        # check behavior if number of samples would exceed file length
        offset_ns = 300
        x_ns, _ = torchaudio.load(input_sine_path, num_frames=100000, offset=offset_ns)
David Pollack's avatar
David Pollack committed
159
        self.assertEqual(x_ns.size(1), x_sine_full.size(1) - offset_ns)
160
161
162
163
164

        # check when offset is beyond the end of the file
        with self.assertRaises(RuntimeError):
            torchaudio.load(input_sine_path, offset=100000)

David Pollack's avatar
David Pollack committed
165
    def test_5_get_info(self):
166
        input_path = os.path.join(self.test_dirpath, 'assets', 'sinewave.wav')
David Pollack's avatar
David Pollack committed
167
168
169
170
171
172
        channels, samples, rate, precision = (1, 64000, 16000, 16)
        si, ei = torchaudio.info(input_path)
        self.assertEqual(si.channels, channels)
        self.assertEqual(si.length, samples)
        self.assertEqual(si.rate, rate)
        self.assertEqual(ei.bits_per_sample, precision)
Soumith Chintala's avatar
Soumith Chintala committed
173

David Pollack's avatar
David Pollack committed
174
175
if __name__ == '__main__':
    unittest.main()