forked from PlayVoice/whisper-vits-svc
-
Notifications
You must be signed in to change notification settings - Fork 0
/
data_utils.py
152 lines (124 loc) · 5.8 KB
/
data_utils.py
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
import time
import os
import random
import numpy as np
import torch
import torch.utils.data
import commons
from mel_processing import spectrogram_torch, spec_to_mel_torch
from utils import load_wav_to_torch, load_filepaths_and_text, transform
# import h5py
"""Multi speaker version"""
class TextAudioSpeakerLoader(torch.utils.data.Dataset):
"""
1) loads audio, speaker_id, text pairs
2) normalizes text and converts them to sequences of integers
3) computes spectrograms from audio files.
"""
def __init__(self, audiopaths, hparams):
self.audiopaths = load_filepaths_and_text(audiopaths)
self.max_wav_value = hparams.data.max_wav_value
self.sampling_rate = hparams.data.sampling_rate
self.filter_length = hparams.data.filter_length
self.hop_length = hparams.data.hop_length
self.win_length = hparams.data.win_length
self.sampling_rate = hparams.data.sampling_rate
self.use_sr = hparams.train.use_sr
self.spec_len = hparams.train.max_speclen
self.spk_map = hparams.spk
random.seed(1234)
random.shuffle(self.audiopaths)
def get_audio(self, filename):
audio, sampling_rate = load_wav_to_torch(filename)
if sampling_rate != self.sampling_rate:
raise ValueError("{} SR doesn't match target {} SR".format(
sampling_rate, self.sampling_rate))
audio_norm = audio / self.max_wav_value
audio_norm = audio_norm.unsqueeze(0)
spec_filename = filename.replace(".wav", ".spec.pt")
if os.path.exists(spec_filename):
spec = torch.load(spec_filename)
else:
spec = spectrogram_torch(audio_norm, self.filter_length,
self.sampling_rate, self.hop_length, self.win_length,
center=False)
spec = torch.squeeze(spec, 0)
torch.save(spec, spec_filename)
spk = filename.split(os.sep)[-2]
spk = torch.LongTensor([self.spk_map[spk]])
c = torch.load(filename + ".soft.pt").squeeze(0)
c = torch.repeat_interleave(c, repeats=2, dim=1)
f0 = np.load(filename + ".f0.npy")
f0 = torch.FloatTensor(f0)
lmin = min(c.size(-1), spec.size(-1), f0.shape[0])
assert abs(c.size(-1) - spec.size(-1)) < 4, (c.size(-1), spec.size(-1), f0.shape, filename)
assert abs(lmin - spec.size(-1)) < 4, (c.size(-1), spec.size(-1), f0.shape)
assert abs(lmin - c.size(-1)) < 4, (c.size(-1), spec.size(-1), f0.shape)
spec, c, f0 = spec[:, :lmin], c[:, :lmin], f0[:lmin]
audio_norm = audio_norm[:, :lmin * self.hop_length]
_spec, _c, _audio_norm, _f0 = spec, c, audio_norm, f0
while spec.size(-1) < self.spec_len:
spec = torch.cat((spec, _spec), -1)
c = torch.cat((c, _c), -1)
f0 = torch.cat((f0, _f0), -1)
audio_norm = torch.cat((audio_norm, _audio_norm), -1)
start = random.randint(0, spec.size(-1) - self.spec_len)
end = start + self.spec_len
spec = spec[:, start:end]
c = c[:, start:end]
f0 = f0[start:end]
audio_norm = audio_norm[:, start * self.hop_length:end * self.hop_length]
return c, f0, spec, audio_norm, spk
def __getitem__(self, index):
return self.get_audio(self.audiopaths[index][0])
def __len__(self):
return len(self.audiopaths)
class EvalDataLoader(torch.utils.data.Dataset):
"""
1) loads audio, speaker_id, text pairs
2) normalizes text and converts them to sequences of integers
3) computes spectrograms from audio files.
"""
def __init__(self, audiopaths, hparams):
self.audiopaths = load_filepaths_and_text(audiopaths)
self.max_wav_value = hparams.data.max_wav_value
self.sampling_rate = hparams.data.sampling_rate
self.filter_length = hparams.data.filter_length
self.hop_length = hparams.data.hop_length
self.win_length = hparams.data.win_length
self.sampling_rate = hparams.data.sampling_rate
self.use_sr = hparams.train.use_sr
self.audiopaths = self.audiopaths[:5]
self.spk_map = hparams.spk
def get_audio(self, filename):
audio, sampling_rate = load_wav_to_torch(filename)
if sampling_rate != self.sampling_rate:
raise ValueError("{} SR doesn't match target {} SR".format(
sampling_rate, self.sampling_rate))
audio_norm = audio / self.max_wav_value
audio_norm = audio_norm.unsqueeze(0)
spec_filename = filename.replace(".wav", ".spec.pt")
if os.path.exists(spec_filename):
spec = torch.load(spec_filename)
else:
spec = spectrogram_torch(audio_norm, self.filter_length,
self.sampling_rate, self.hop_length, self.win_length,
center=False)
spec = torch.squeeze(spec, 0)
torch.save(spec, spec_filename)
spk = filename.split(os.sep)[-2]
spk = torch.LongTensor([self.spk_map[spk]])
c = torch.load(filename + ".soft.pt").squeeze(0)
c = torch.repeat_interleave(c, repeats=2, dim=1)
f0 = np.load(filename + ".f0.npy")
f0 = torch.FloatTensor(f0)
lmin = min(c.size(-1), spec.size(-1), f0.shape[0])
assert abs(c.size(-1) - spec.size(-1)) < 4, (c.size(-1), spec.size(-1), f0.shape)
assert abs(f0.shape[0] - spec.shape[-1]) < 4, (c.size(-1), spec.size(-1), f0.shape)
spec, c, f0 = spec[:, :lmin], c[:, :lmin], f0[:lmin]
audio_norm = audio_norm[:, :lmin * self.hop_length]
return c, f0, spec, audio_norm, spk
def __getitem__(self, index):
return self.get_audio(self.audiopaths[index][0])
def __len__(self):
return len(self.audiopaths)