wassname/TTS
1
0
Fork 0
mirror of https://github.com/wassname/TTS.git synced 2026-09-09 11:16:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

pep8 format all

Browse Source
This commit is contained in:
Eren G
2018-08-02 16:34:17 +02:00
parent 3238ffa3e6
commit f5537dc48f
32 changed files with 766 additions and 599 deletions
Download Patch File Download Diff File Expand all files Collapse all files
utils/audio.py
+23 -11
View File
@@ -9,10 +9,19 @@ _mel_basis = None
class AudioProcessor(object):
def __init__(self, sample_rate, num_mels, min_level_db, frame_shift_ms,
frame_length_ms, ref_level_db, num_freq, power, preemphasis,
min_mel_freq, max_mel_freq, griffin_lim_iters=None):
def __init__(self,
sample_rate,
num_mels,
min_level_db,
frame_shift_ms,
frame_length_ms,
ref_level_db,
num_freq,
power,
preemphasis,
min_mel_freq,
max_mel_freq,
griffin_lim_iters=None):
self.sample_rate = sample_rate
self.num_mels = num_mels
@@ -30,7 +39,8 @@ class AudioProcessor(object):
def save_wav(self, wav, path):
wav *= 32767 / max(0.01, np.max(np.abs(wav)))
librosa.output.write_wav(path, wav.astype(np.float), self.sample_rate, norm=True)
librosa.output.write_wav(
path, wav.astype(np.float), self.sample_rate, norm=True)
def _linear_to_mel(self, spectrogram):
global _mel_basis
@@ -40,8 +50,9 @@ class AudioProcessor(object):
def _build_mel_basis(self, ):
n_fft = (self.num_freq - 1) * 2
return librosa.filters.mel(self.sample_rate, n_fft, n_mels=self.num_mels)
# fmin=self.min_mel_freq, fmax=self.max_mel_freq)
return librosa.filters.mel(
self.sample_rate, n_fft, n_mels=self.num_mels)
# fmin=self.min_mel_freq, fmax=self.max_mel_freq)
def _normalize(self, S):
return np.clip((S - self.min_level_db) / -self.min_level_db, 0, 1)
@@ -66,7 +77,7 @@ class AudioProcessor(object):
if self.preemphasis == 0:
raise RuntimeError(" !! Preemphasis is applied with factor 0.0. ")
return signal.lfilter([1, -self.preemphasis], [1], x)
def apply_inv_preemphasis(self, x):
if self.preemphasis == 0:
raise RuntimeError(" !! Preemphasis is applied with factor 0.0. ")
@@ -86,9 +97,9 @@ class AudioProcessor(object):
S = self._db_to_amp(S + self.ref_level_db) # Convert back to linear
# Reconstruct phase
if self.preemphasis != 0:
return self.apply_inv_preemphasis(self._griffin_lim(S ** self.power))
return self.apply_inv_preemphasis(self._griffin_lim(S**self.power))
else:
return self._griffin_lim(S ** self.power)
return self._griffin_lim(S**self.power)
def _griffin_lim(self, S):
'''Applies Griffin-Lim's raw.
@@ -113,7 +124,8 @@ class AudioProcessor(object):
def _stft(self, y):
n_fft, hop_length, win_length = self._stft_parameters()
return librosa.stft(y=y, n_fft=n_fft, hop_length=hop_length, win_length=win_length)
return librosa.stft(
y=y, n_fft=n_fft, hop_length=hop_length, win_length=win_length)
def _istft(self, y):
_, hop_length, win_length = self._stft_parameters()
utils/audio_lws.py
+40 -17
View File
@@ -8,11 +8,23 @@ import lws
_mel_basis = None
class AudioProcessor(object):
def __init__(self, sample_rate, num_mels, min_level_db, frame_shift_ms,
frame_length_ms, ref_level_db, num_freq, power, preemphasis,
min_mel_freq, max_mel_freq, griffin_lim_iters=None, ):
class AudioProcessor(object):
def __init__(
self,
sample_rate,
num_mels,
min_level_db,
frame_shift_ms,
frame_length_ms,
ref_level_db,
num_freq,
power,
preemphasis,
min_mel_freq,
max_mel_freq,
griffin_lim_iters=None,
):
print(" > Setting up Audio Processor...")
self.sample_rate = sample_rate
self.num_mels = num_mels
@@ -25,18 +37,19 @@ class AudioProcessor(object):
self.min_mel_freq = min_mel_freq
self.max_mel_freq = max_mel_freq
self.griffin_lim_iters = griffin_lim_iters
self.preemphasis =preemphasis
self.preemphasis = preemphasis
self.n_fft, self.hop_length, self.win_length = self._stft_parameters()
if preemphasis == 0:
print(" | > Preemphasis is deactive.")
def save_wav(self, wav, path):
wav *= 32767 / max(0.01, np.max(np.abs(wav)))
librosa.output.write_wav(path, wav.astype(np.float), self.sample_rate, norm=True)
librosa.output.write_wav(
path, wav.astype(np.float), self.sample_rate, norm=True)
def _stft_parameters(self, ):
n_fft = int((self.num_freq - 1) * 2)
hop_length = int(self.frame_shift_ms / 1000.0 * self.sample_rate)
hop_length = int(self.frame_shift_ms / 1000.0 * self.sample_rate)
win_length = int(self.frame_length_ms / 1000.0 * self.sample_rate)
if n_fft % hop_length != 0:
hop_length = n_fft / 8
@@ -44,14 +57,21 @@ class AudioProcessor(object):
if n_fft % win_length != 0:
win_length = n_fft / 2
print(" | > win_length is set to default ({}).".format(win_length))
print(" | > fft size: {}, hop length: {}, win length: {}".format(n_fft, hop_length, win_length))
print(" | > fft size: {}, hop length: {}, win length: {}".format(
n_fft, hop_length, win_length))
return int(n_fft), int(hop_length), int(win_length)
def _lws_processor(self):
try:
return lws.lws(self.win_length, self.hop_length, fftsize=self.n_fft, mode="speech")
return lws.lws(
self.win_length,
self.hop_length,
fftsize=self.n_fft,
mode="speech")
except:
raise RuntimeError(" !! WindowLength({}) is not multiple of HopLength({}).".format(self.win_length, self.hop_length))
raise RuntimeError(
" !! WindowLength({}) is not multiple of HopLength({}).".
format(self.win_length, self.hop_length))
def _amp_to_db(self, x):
min_level = np.exp(self.min_level_db / 20 * np.log(10))
@@ -70,7 +90,7 @@ class AudioProcessor(object):
if self.preemphasis == 0:
raise RuntimeError(" !! Preemphasis is applied with factor 0.0. ")
return signal.lfilter([1, -self.preemphasis], [1], x)
def apply_inv_preemphasis(self, x):
if self.preemphasis == 0:
raise RuntimeError(" !! Preemphasis is applied with factor 0.0. ")
@@ -96,14 +116,14 @@ class AudioProcessor(object):
S = self._denormalize(spectrogram)
S = self._db_to_amp(S + self.ref_level_db) # Convert back to linear
processor = self._lws_processor()
D = processor.run_lws(S.astype(np.float64).T ** self.power)
D = processor.run_lws(S.astype(np.float64).T**self.power)
y = processor.istft(D).astype(np.float32)
# Reconstruct phase
if self.preemphasis:
return self.apply_inv_preemphasis(y)
sys.stdout = old_out
return y
def _linear_to_mel(self, spectrogram):
global _mel_basis
if _mel_basis is None:
@@ -111,7 +131,10 @@ class AudioProcessor(object):
return np.dot(_mel_basis, spectrogram)
def _build_mel_basis(self, ):
return librosa.filters.mel(self.sample_rate, self.n_fft, n_mels=self.num_mels)
return librosa.filters.mel(
self.sample_rate, self.n_fft, n_mels=self.num_mels)
# fmin=self.min_mel_freq, fmax=self.max_mel_freq)
def melspectrogram(self, y):
@@ -124,4 +147,4 @@ class AudioProcessor(object):
D = self._lws_processor().stft(y).T
S = self._amp_to_db(self._linear_to_mel(np.abs(D))) - self.ref_level_db
sys.stdout = old_out
return self._normalize(S)
return self._normalize(S)
utils/data.py
+12 -9
View File
@@ -4,9 +4,8 @@ import numpy as np
def _pad_data(x, length):
_pad = 0
assert x.ndim == 1
return np.pad(x, (0, length - x.shape[0]),
mode='constant',
constant_values=_pad)
return np.pad(
x, (0, length - x.shape[0]), mode='constant', constant_values=_pad)
def prepare_data(inputs):
@@ -17,8 +16,10 @@ def prepare_data(inputs):
def _pad_tensor(x, length):
_pad = 0
assert x.ndim == 2
x = np.pad(x, [[0, 0], [0, length - x.shape[1]]],
mode='constant', constant_values=_pad)
x = np.pad(
x, [[0, 0], [0, length - x.shape[1]]],
mode='constant',
constant_values=_pad)
return x
@@ -32,7 +33,8 @@ def prepare_tensor(inputs, out_steps):
def _pad_stop_target(x, length):
_pad = 1.
assert x.ndim == 1
return np.pad(x, (0, length - x.shape[0]), mode='constant', constant_values=_pad)
return np.pad(
x, (0, length - x.shape[0]), mode='constant', constant_values=_pad)
def prepare_stop_target(inputs, out_steps):
@@ -44,6 +46,7 @@ def prepare_stop_target(inputs, out_steps):
def pad_per_step(inputs, pad_len):
timesteps = inputs.shape[-1]
return np.pad(inputs, [[0, 0], [0, 0],
[0, pad_len]],
mode='constant', constant_values=0.0)
return np.pad(
inputs, [[0, 0], [0, 0], [0, pad_len]],
mode='constant',
constant_values=0.0)
utils/generic_utils.py
+35 -27
View File
@@ -28,10 +28,13 @@ def load_config(config_path):
def get_commit_hash():
"""https://stackoverflow.com/questions/14989858/get-the-current-git-hash-in-a-python-script"""
try:
subprocess.check_output(['git', 'diff-index', '--quiet', 'HEAD']) # Verify client is clean
subprocess.check_output(['git', 'diff-index', '--quiet',
'HEAD']) # Verify client is clean
except:
raise RuntimeError(" !! Commit before training to get the commit hash.")
commit = subprocess.check_output(['git', 'rev-parse', '--short', 'HEAD']).decode().strip()
raise RuntimeError(
" !! Commit before training to get the commit hash.")
commit = subprocess.check_output(['git', 'rev-parse', '--short',
'HEAD']).decode().strip()
print(' > Git Hash: {}'.format(commit))
return commit
@@ -43,7 +46,8 @@ def create_experiment_folder(root_path, model_name, debug):
commit_hash = 'debug'
else:
commit_hash = get_commit_hash()
output_folder = os.path.join(root_path, date_str + '-' + model_name + '-' + commit_hash)
output_folder = os.path.join(
root_path, date_str + '-' + model_name + '-' + commit_hash)
os.makedirs(output_folder, exist_ok=True)
print(" > Experiment folder: {}".format(output_folder))
return output_folder
@@ -52,7 +56,7 @@ def create_experiment_folder(root_path, model_name, debug):
def remove_experiment_folder(experiment_path):
"""Check folder if there is a checkpoint, otherwise remove the folder"""
checkpoint_files = glob.glob(experiment_path+"/*.pth.tar")
checkpoint_files = glob.glob(experiment_path + "/*.pth.tar")
if len(checkpoint_files) < 1:
if os.path.exists(experiment_path):
shutil.rmtree(experiment_path)
@@ -86,13 +90,15 @@ def save_checkpoint(model, optimizer, optimizer_st, model_loss, out_path,
print(" | | > Checkpoint saving : {}".format(checkpoint_path))
new_state_dict = _trim_model_state_dict(model.state_dict())
state = {'model': new_state_dict,
'optimizer': optimizer.state_dict(),
'optimizer_st': optimizer_st.state_dict(),
'step': current_step,
'epoch': epoch,
'linear_loss': model_loss,
'date': datetime.date.today().strftime("%B %d, %Y")}
state = {
'model': new_state_dict,
'optimizer': optimizer.state_dict(),
'optimizer_st': optimizer_st.state_dict(),
'step': current_step,
'epoch': epoch,
'linear_loss': model_loss,
'date': datetime.date.today().strftime("%B %d, %Y")
}
torch.save(state, checkpoint_path)
@@ -100,12 +106,14 @@ def save_best_model(model, optimizer, model_loss, best_loss, out_path,
current_step, epoch):
if model_loss < best_loss:
new_state_dict = _trim_model_state_dict(model.state_dict())
state = {'model': new_state_dict,
'optimizer': optimizer.state_dict(),
'step': current_step,
'epoch': epoch,
'linear_loss': model_loss,
'date': datetime.date.today().strftime("%B %d, %Y")}
state = {
'model': new_state_dict,
'optimizer': optimizer.state_dict(),
'step': current_step,
'epoch': epoch,
'linear_loss': model_loss,
'date': datetime.date.today().strftime("%B %d, %Y")
}
best_loss = model_loss
bestmodel_path = 'best_model.pth.tar'
bestmodel_path = os.path.join(out_path, bestmodel_path)
@@ -161,12 +169,12 @@ def sequence_mask(sequence_length, max_len=None):
def synthesis(model, ap, text, use_cuda, text_cleaner):
text_cleaner = [text_cleaner]
seq = np.array(text_to_sequence(text, text_cleaner))
chars_var = torch.from_numpy(seq).unsqueeze(0)
if use_cuda:
chars_var = chars_var.cuda().long()
_, linear_out, alignments, _ = model.forward(chars_var)
linear_out = linear_out[0].data.cpu().numpy()
wav = ap.inv_spectrogram(linear_out.T)
return wav, linear_out, alignments
text_cleaner = [text_cleaner]
seq = np.array(text_to_sequence(text, text_cleaner))
chars_var = torch.from_numpy(seq).unsqueeze(0)
if use_cuda:
chars_var = chars_var.cuda().long()
_, linear_out, alignments, _ = model.forward(chars_var)
linear_out = linear_out[0].data.cpu().numpy()
wav = ap.inv_spectrogram(linear_out.T)
return wav, linear_out, alignments
utils/text/__init__.py
-1
View File
@@ -4,7 +4,6 @@ import re
from utils.text import cleaners
from utils.text.symbols import symbols
# Mappings from symbol to numeric ID and vice versa:
_symbol_to_id = {s: i for i, s in enumerate(symbols)}
_id_to_symbol = {i: s for i, s in enumerate(symbols)}
utils/text/cleaners.py
+21 -21
View File
@@ -14,31 +14,31 @@ import re
from unidecode import unidecode
from .numbers import normalize_numbers
# Regular expression matching whitespace:
_whitespace_re = re.compile(r'\s+')
# List of (regular expression, replacement) pairs for abbreviations:
_abbreviations = [(re.compile('\\b%s\\.' % x[0], re.IGNORECASE), x[1]) for x in [
('mrs', 'misess'),
('mr', 'mister'),
('dr', 'doctor'),
('st', 'saint'),
('co', 'company'),
('jr', 'junior'),
('maj', 'major'),
('gen', 'general'),
('drs', 'doctors'),
('rev', 'reverend'),
('lt', 'lieutenant'),
('hon', 'honorable'),
('sgt', 'sergeant'),
('capt', 'captain'),
('esq', 'esquire'),
('ltd', 'limited'),
('col', 'colonel'),
('ft', 'fort'),
]]
_abbreviations = [(re.compile('\\b%s\\.' % x[0], re.IGNORECASE), x[1])
for x in [
('mrs', 'misess'),
('mr', 'mister'),
('dr', 'doctor'),
('st', 'saint'),
('co', 'company'),
('jr', 'junior'),
('maj', 'major'),
('gen', 'general'),
('drs', 'doctors'),
('rev', 'reverend'),
('lt', 'lieutenant'),
('hon', 'honorable'),
('sgt', 'sergeant'),
('capt', 'captain'),
('esq', 'esquire'),
('ltd', 'limited'),
('col', 'colonel'),
('ft', 'fort'),
]]
def expand_abbreviations(text):
utils/text/cmudict.py
+12 -11
View File
@@ -1,17 +1,16 @@
# -*- coding: utf-8 -*-
import re
valid_symbols = [
'AA', 'AA0', 'AA1', 'AA2', 'AE', 'AE0', 'AE1', 'AE2', 'AH', 'AH0', 'AH1', 'AH2',
'AO', 'AO0', 'AO1', 'AO2', 'AW', 'AW0', 'AW1', 'AW2', 'AY', 'AY0', 'AY1', 'AY2',
'B', 'CH', 'D', 'DH', 'EH', 'EH0', 'EH1', 'EH2', 'ER', 'ER0', 'ER1', 'ER2', 'EY',
'EY0', 'EY1', 'EY2', 'F', 'G', 'HH', 'IH', 'IH0', 'IH1', 'IH2', 'IY', 'IY0', 'IY1',
'IY2', 'JH', 'K', 'L', 'M', 'N', 'NG', 'OW', 'OW0', 'OW1', 'OW2', 'OY', 'OY0',
'OY1', 'OY2', 'P', 'R', 'S', 'SH', 'T', 'TH', 'UH', 'UH0', 'UH1', 'UH2', 'UW',
'UW0', 'UW1', 'UW2', 'V', 'W', 'Y', 'Z', 'ZH'
'AA', 'AA0', 'AA1', 'AA2', 'AE', 'AE0', 'AE1', 'AE2', 'AH', 'AH0', 'AH1',
'AH2', 'AO', 'AO0', 'AO1', 'AO2', 'AW', 'AW0', 'AW1', 'AW2', 'AY', 'AY0',
'AY1', 'AY2', 'B', 'CH', 'D', 'DH', 'EH', 'EH0', 'EH1', 'EH2', 'ER', 'ER0',
'ER1', 'ER2', 'EY', 'EY0', 'EY1', 'EY2', 'F', 'G', 'HH', 'IH', 'IH0',
'IH1', 'IH2', 'IY', 'IY0', 'IY1', 'IY2', 'JH', 'K', 'L', 'M', 'N', 'NG',
'OW', 'OW0', 'OW1', 'OW2', 'OY', 'OY0', 'OY1', 'OY2', 'P', 'R', 'S', 'SH',
'T', 'TH', 'UH', 'UH0', 'UH1', 'UH2', 'UW', 'UW0', 'UW1', 'UW2', 'V', 'W',
'Y', 'Z', 'ZH'
]
_valid_symbol_set = set(valid_symbols)
@@ -27,8 +26,10 @@ class CMUDict:
else:
entries = _parse_cmudict(file_or_path)
if not keep_ambiguous:
entries = {word: pron for word,
pron in entries.items() if len(pron) == 1}
entries = {
word: pron
for word, pron in entries.items() if len(pron) == 1
}
self._entries = entries
def __len__(self):
utils/text/numbers.py
+29 -45
View File
@@ -8,61 +8,45 @@ _ordinal_re = re.compile(r'([0-9]+)(st|nd|rd|th)')
_number_re = re.compile(r'[0-9]+')
_units = [
'',
'one',
'two',
'three',
'four',
'five',
'six',
'seven',
'eight',
'nine',
'ten',
'eleven',
'twelve',
'thirteen',
'fourteen',
'fifteen',
'sixteen',
'seventeen',
'eighteen',
'nineteen'
'', 'one', 'two', 'three', 'four', 'five', 'six', 'seven', 'eight', 'nine',
'ten', 'eleven', 'twelve', 'thirteen', 'fourteen', 'fifteen', 'sixteen',
'seventeen', 'eighteen', 'nineteen'
]
_tens = [
'',
'ten',
'twenty',
'thirty',
'forty',
'fifty',
'sixty',
'seventy',
'eighty',
'ninety',
'',
'ten',
'twenty',
'thirty',
'forty',
'fifty',
'sixty',
'seventy',
'eighty',
'ninety',
]
_digit_groups = [
'',
'thousand',
'million',
'billion',
'trillion',
'quadrillion',
'',
'thousand',
'million',
'billion',
'trillion',
'quadrillion',
]
_ordinal_suffixes = [
('one', 'first'),
('two', 'second'),
('three', 'third'),
('five', 'fifth'),
('eight', 'eighth'),
('nine', 'ninth'),
('twelve', 'twelfth'),
('ty', 'tieth'),
('one', 'first'),
('two', 'second'),
('three', 'third'),
('five', 'fifth'),
('eight', 'eighth'),
('nine', 'ninth'),
('twelve', 'twelfth'),
('ty', 'tieth'),
]
def _remove_commas(m):
return m.group(1).replace(',', '')
@@ -114,7 +98,7 @@ def _standard_number_to_words(n, digit_group):
def _number_to_words(n):
# Handle special cases first, then go to the standard case:
if n >= 1000000000000000000:
return str(n) # Too large, just return the digits
return str(n) # Too large, just return the digits
elif n == 0:
return 'zero'
elif n % 100 == 0 and n % 1000 != 0 and n < 3000:
utils/text/symbols.py
-3
View File
@@ -1,6 +1,4 @@
# -*- coding: utf-8 -*-
'''
Defines the set of symbols used in text input to the model.
@@ -19,6 +17,5 @@ _arpabet = ['@' + s for s in cmudict.valid_symbols]
# Export all symbols:
symbols = [_pad, _eos] + list(_characters) + _arpabet
if __name__ == '__main__':
print(symbols)
utils/visual.py
+5 -5
View File
@@ -6,8 +6,8 @@ import matplotlib.pyplot as plt
def plot_alignment(alignment, info=None):
fig, ax = plt.subplots(figsize=(16, 10))
im = ax.imshow(alignment.T, aspect='auto', origin='lower',
interpolation='none')
im = ax.imshow(
alignment.T, aspect='auto', origin='lower', interpolation='none')
fig.colorbar(im, ax=ax)
xlabel = 'Decoder timestep'
if info is not None:
@@ -17,7 +17,7 @@ def plot_alignment(alignment, info=None):
plt.tight_layout()
fig.canvas.draw()
data = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep='')
data = data.reshape(fig.canvas.get_width_height()[::-1] + (3,))
data = data.reshape(fig.canvas.get_width_height()[::-1] + (3, ))
plt.close()
return data
@@ -30,6 +30,6 @@ def plot_spectrogram(linear_output, audio):
plt.tight_layout()
fig.canvas.draw()
data = np.fromstring(fig.canvas.tostring_rgb(), dtype=np.uint8, sep='')
data = data.reshape(fig.canvas.get_width_height()[::-1] + (3,))
data = data.reshape(fig.canvas.get_width_height()[::-1] + (3, ))
plt.close()
return data
return data