Spaces:
Runtime error
Runtime error
File size: 10,428 Bytes
3b92d66 |
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 |
import random
import numpy as np
import torch
import torch.utils.data
import commons
from utils import load_wav_to_torch, load_filepaths_and_text
from text import text_to_sequence
class TextMelLoader(torch.utils.data.Dataset):
"""
1) loads audio,text pairs
2) normalizes text and converts them to sequences of one-hot vectors
3) computes mel-spectrograms from audio files.
"""
def __init__(self, audiopaths_and_text, hparams):
self.audiopaths_and_text = load_filepaths_and_text(audiopaths_and_text)
self.text_cleaners = hparams.text_cleaners
self.max_wav_value = hparams.max_wav_value
self.sampling_rate = hparams.sampling_rate
self.load_mel_from_disk = hparams.load_mel_from_disk
self.add_noise = hparams.add_noise
self.symbols = hparams.punc + hparams.chars
self.add_blank = getattr(hparams, "add_blank", False) # improved version
self.stft = commons.TacotronSTFT(
hparams.filter_length,
hparams.hop_length,
hparams.win_length,
hparams.n_mel_channels,
hparams.sampling_rate,
hparams.mel_fmin,
hparams.mel_fmax,
)
random.seed(1234)
random.shuffle(self.audiopaths_and_text)
def get_mel_text_pair(self, audiopath_and_text):
# separate filename and text
audiopath, text = audiopath_and_text[0], audiopath_and_text[1]
text = self.get_text(text)
mel = self.get_mel(audiopath)
return (text, mel)
def get_mel(self, filename):
if not self.load_mel_from_disk:
audio, sampling_rate = load_wav_to_torch(filename)
if sampling_rate != self.stft.sampling_rate:
raise ValueError(
"{} {} SR doesn't match target {} SR".format(
sampling_rate, self.stft.sampling_rate
)
)
if self.add_noise:
audio = audio + torch.rand_like(audio)
audio_norm = audio / self.max_wav_value
audio_norm = audio_norm.unsqueeze(0)
melspec = self.stft.mel_spectrogram(audio_norm)
melspec = torch.squeeze(melspec, 0)
else:
melspec = torch.from_numpy(np.load(filename))
assert (
melspec.size(0) == self.stft.n_mel_channels
), "Mel dimension mismatch: given {}, expected {}".format(
melspec.size(0), self.stft.n_mel_channels
)
return melspec
def get_text(self, text):
text_norm = text_to_sequence(text, self.symbols, self.text_cleaners)
if self.add_blank:
text_norm = commons.intersperse(
text_norm, len(self.symbols)
) # add a blank token, whose id number is len(symbols)
text_norm = torch.IntTensor(text_norm)
return text_norm
def __getitem__(self, index):
return self.get_mel_text_pair(self.audiopaths_and_text[index])
def __len__(self):
return len(self.audiopaths_and_text)
class TextMelCollate:
"""Zero-pads model inputs and targets based on number of frames per step"""
def __init__(self, n_frames_per_step=1):
self.n_frames_per_step = n_frames_per_step
def __call__(self, batch):
"""Collate's training batch from normalized text and mel-spectrogram
PARAMS
------
batch: [text_normalized, mel_normalized]
"""
# Right zero-pad all one-hot text sequences to max input length
input_lengths, ids_sorted_decreasing = torch.sort(
torch.LongTensor([len(x[0]) for x in batch]), dim=0, descending=True
)
max_input_len = input_lengths[0]
text_padded = torch.LongTensor(len(batch), max_input_len)
text_padded.zero_()
for i in range(len(ids_sorted_decreasing)):
text = batch[ids_sorted_decreasing[i]][0]
text_padded[i, : text.size(0)] = text
# Right zero-pad mel-spec
num_mels = batch[0][1].size(0)
max_target_len = max([x[1].size(1) for x in batch])
if max_target_len % self.n_frames_per_step != 0:
max_target_len += (
self.n_frames_per_step - max_target_len % self.n_frames_per_step
)
assert max_target_len % self.n_frames_per_step == 0
# include mel padded
mel_padded = torch.FloatTensor(len(batch), num_mels, max_target_len)
mel_padded.zero_()
output_lengths = torch.LongTensor(len(batch))
for i in range(len(ids_sorted_decreasing)):
mel = batch[ids_sorted_decreasing[i]][1]
mel_padded[i, :, : mel.size(1)] = mel
output_lengths[i] = mel.size(1)
return text_padded, input_lengths, mel_padded, output_lengths
"""Multi speaker version"""
class TextMelSpeakerLoader(torch.utils.data.Dataset):
"""
1) loads audio, speaker_id, text pairs
2) normalizes text and converts them to sequences of one-hot vectors
3) computes mel-spectrograms from audio files.
"""
def __init__(self, audiopaths_sid_text, hparams):
self.audiopaths_sid_text = load_filepaths_and_text(audiopaths_sid_text)
self.text_cleaners = hparams.text_cleaners
self.max_wav_value = hparams.max_wav_value
self.sampling_rate = hparams.sampling_rate
self.load_mel_from_disk = hparams.load_mel_from_disk
self.add_noise = hparams.add_noise
self.symbols = hparams.punc + hparams.chars
self.add_blank = getattr(hparams, "add_blank", False) # improved version
self.min_text_len = getattr(hparams, "min_text_len", 1)
self.max_text_len = getattr(hparams, "max_text_len", 190)
self.stft = commons.TacotronSTFT(
hparams.filter_length,
hparams.hop_length,
hparams.win_length,
hparams.n_mel_channels,
hparams.sampling_rate,
hparams.mel_fmin,
hparams.mel_fmax,
)
self._filter_text_len()
random.seed(1234)
random.shuffle(self.audiopaths_sid_text)
def _filter_text_len(self):
audiopaths_sid_text_new = []
for audiopath, sid, text in self.audiopaths_sid_text:
if self.min_text_len <= len(text) and len(text) <= self.max_text_len:
audiopaths_sid_text_new.append([audiopath, sid, text])
self.audiopaths_sid_text = audiopaths_sid_text_new
def get_mel_text_speaker_pair(self, audiopath_sid_text):
# separate filename, speaker_id and text
audiopath, sid, text = (
audiopath_sid_text[0],
audiopath_sid_text[1],
audiopath_sid_text[2],
)
text = self.get_text(text)
mel = self.get_mel(audiopath)
sid = self.get_sid(sid)
return (text, mel, sid)
def get_mel(self, filename):
if not self.load_mel_from_disk:
audio, sampling_rate = load_wav_to_torch(filename)
if sampling_rate != self.stft.sampling_rate:
raise ValueError(
"{} {} SR doesn't match target {} SR".format(
sampling_rate, self.stft.sampling_rate
)
)
if self.add_noise:
audio = audio + torch.rand_like(audio)
audio_norm = audio / self.max_wav_value
audio_norm = audio_norm.unsqueeze(0)
melspec = self.stft.mel_spectrogram(audio_norm)
melspec = torch.squeeze(melspec, 0)
else:
melspec = torch.from_numpy(np.load(filename))
assert (
melspec.size(0) == self.stft.n_mel_channels
), "Mel dimension mismatch: given {}, expected {}".format(
melspec.size(0), self.stft.n_mel_channels
)
return melspec
def get_text(self, text):
text_norm = text_to_sequence(text, self.symbols, self.text_cleaners)
if self.add_blank:
text_norm = commons.intersperse(
text_norm, len(self.symbols)
) # add a blank token, whose id number is len(symbols)
text_norm = torch.IntTensor(text_norm)
return text_norm
def get_sid(self, sid):
sid = torch.IntTensor([int(sid)])
return sid
def __getitem__(self, index):
return self.get_mel_text_speaker_pair(self.audiopaths_sid_text[index])
def __len__(self):
return len(self.audiopaths_sid_text)
class TextMelSpeakerCollate:
"""Zero-pads model inputs and targets based on number of frames per step"""
def __init__(self, n_frames_per_step=1):
self.n_frames_per_step = n_frames_per_step
def __call__(self, batch):
"""Collate's training batch from normalized text and mel-spectrogram
PARAMS
------
batch: [text_normalized, mel_normalized]
"""
# Right zero-pad all one-hot text sequences to max input length
input_lengths, ids_sorted_decreasing = torch.sort(
torch.LongTensor([len(x[0]) for x in batch]), dim=0, descending=True
)
max_input_len = input_lengths[0]
text_padded = torch.LongTensor(len(batch), max_input_len)
text_padded.zero_()
for i in range(len(ids_sorted_decreasing)):
text = batch[ids_sorted_decreasing[i]][0]
text_padded[i, : text.size(0)] = text
# Right zero-pad mel-spec
num_mels = batch[0][1].size(0)
max_target_len = max([x[1].size(1) for x in batch])
if max_target_len % self.n_frames_per_step != 0:
max_target_len += (
self.n_frames_per_step - max_target_len % self.n_frames_per_step
)
assert max_target_len % self.n_frames_per_step == 0
# include mel padded & sid
mel_padded = torch.FloatTensor(len(batch), num_mels, max_target_len)
mel_padded.zero_()
output_lengths = torch.LongTensor(len(batch))
sid = torch.LongTensor(len(batch))
for i in range(len(ids_sorted_decreasing)):
mel = batch[ids_sorted_decreasing[i]][1]
mel_padded[i, :, : mel.size(1)] = mel
output_lengths[i] = mel.size(1)
sid[i] = batch[ids_sorted_decreasing[i]][2]
return text_padded, input_lengths, mel_padded, output_lengths, sid
|