Medix-AI/coqui-tts
Medix-AI
/
coqui-tts
mirror of https://github.com/coqui-ai/TTS.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #20 from eginhard/return-complex 

fix: torch.stft will soon require return_complex=True
This commit is contained in:
Enno Hermann 2024-03-13 13:50:21 +01:00 committed by GitHub
parent eaa7283244 e95f8950eb
commit e5c6da1c98
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
4 changed files with 91 additions and 77 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

48
TTS/tts/models/delightful_tts.py
View File

@ -179,17 +179,19 @@ def _wav_to_spec(y, n_fft, hop_length, win_length, center=False):
)
y = y.squeeze(1)
spec = torch.stft(
y,
n_fft,
hop_length=hop_length,
win_length=win_length,
window=hann_window[wnsize_dtype_device],
center=center,
pad_mode="reflect",
normalized=False,
onesided=True,
return_complex=False,
spec = torch.view_as_real(
torch.stft(
y,
n_fft,
hop_length=hop_length,
win_length=win_length,
window=hann_window[wnsize_dtype_device],
center=center,
pad_mode="reflect",
normalized=False,
onesided=True,
return_complex=True,
)
)
return spec
@ -274,17 +276,19 @@ def wav_to_mel(y, n_fft, num_mels, sample_rate, hop_length, win_length, fmin, fm
)
y = y.squeeze(1)
spec = torch.stft(
y,
n_fft,
hop_length=hop_length,
win_length=win_length,
window=hann_window[wnsize_dtype_device],
center=center,
pad_mode="reflect",
normalized=False,
onesided=True,
return_complex=False,
spec = torch.view_as_real(
torch.stft(
y,
n_fft,
hop_length=hop_length,
win_length=win_length,
window=hann_window[wnsize_dtype_device],
center=center,
pad_mode="reflect",
normalized=False,
onesided=True,
return_complex=True,
)
)
spec = torch.sqrt(spec.pow(2).sum(-1) + 1e-6)

48
TTS/tts/models/vits.py
View File

@ -121,17 +121,19 @@ def wav_to_spec(y, n_fft, hop_length, win_length, center=False):
)
y = y.squeeze(1)
spec = torch.stft(
y,
n_fft,
hop_length=hop_length,
win_length=win_length,
window=hann_window[wnsize_dtype_device],
center=center,
pad_mode="reflect",
normalized=False,
onesided=True,
return_complex=False,
spec = torch.view_as_real(
torch.stft(
y,
n_fft,
hop_length=hop_length,
win_length=win_length,
window=hann_window[wnsize_dtype_device],
center=center,
pad_mode="reflect",
normalized=False,
onesided=True,
return_complex=True,
)
)
spec = torch.sqrt(spec.pow(2).sum(-1) + 1e-6)
@ -189,17 +191,19 @@ def wav_to_mel(y, n_fft, num_mels, sample_rate, hop_length, win_length, fmin, fm
)
y = y.squeeze(1)
spec = torch.stft(
y,
n_fft,
hop_length=hop_length,
win_length=win_length,
window=hann_window[wnsize_dtype_device],
center=center,
pad_mode="reflect",
normalized=False,
onesided=True,
return_complex=False,
spec = torch.view_as_real(
torch.stft(
y,
n_fft,
hop_length=hop_length,
win_length=win_length,
window=hann_window[wnsize_dtype_device],
center=center,
pad_mode="reflect",
normalized=False,
onesided=True,
return_complex=True,
)
)
spec = torch.sqrt(spec.pow(2).sum(-1) + 1e-6)

24
TTS/utils/audio/torch_transforms.py
View File

@ -119,17 +119,19 @@ class TorchSTFT(nn.Module): # pylint: disable=abstract-method
padding = int((self.n_fft - self.hop_length) / 2)
x = torch.nn.functional.pad(x, (padding, padding), mode="reflect")
# B x D x T x 2
o = torch.stft(
x.squeeze(1),
self.n_fft,
self.hop_length,
self.win_length,
self.window,
center=True,
pad_mode="reflect", # compatible with audio.py
normalized=self.normalized,
onesided=True,
return_complex=False,
o = torch.view_as_real(
torch.stft(
x.squeeze(1),
self.n_fft,
self.hop_length,
self.win_length,
self.window,
center=True,
pad_mode="reflect", # compatible with audio.py
normalized=self.normalized,
onesided=True,
return_complex=True,
)
)
M = o[:, :, :, 0]
P = o[:, :, :, 1]

48
TTS/vc/modules/freevc/mel_processing.py
View File

@ -54,17 +54,19 @@ def spectrogram_torch(y, n_fft, sampling_rate, hop_size, win_size, center=False)
)
y = y.squeeze(1)
spec = torch.stft(
y,
n_fft,
hop_length=hop_size,
win_length=win_size,
window=hann_window[wnsize_dtype_device],
center=center,
pad_mode="reflect",
normalized=False,
onesided=True,
return_complex=False,
spec = torch.view_as_real(
torch.stft(
y,
n_fft,
hop_length=hop_size,
win_length=win_size,
window=hann_window[wnsize_dtype_device],
center=center,
pad_mode="reflect",
normalized=False,
onesided=True,
return_complex=True,
)
)
spec = torch.sqrt(spec.pow(2).sum(-1) + 1e-6)
@ -104,17 +106,19 @@ def mel_spectrogram_torch(y, n_fft, num_mels, sampling_rate, hop_size, win_size,
)
y = y.squeeze(1)
spec = torch.stft(
y,
n_fft,
hop_length=hop_size,
win_length=win_size,
window=hann_window[wnsize_dtype_device],
center=center,
pad_mode="reflect",
normalized=False,
onesided=True,
return_complex=False,
spec = torch.view_as_real(
torch.stft(
y,
n_fft,
hop_length=hop_size,
win_length=win_size,
window=hann_window[wnsize_dtype_device],
center=center,
pad_mode="reflect",
normalized=False,
onesided=True,
return_complex=True,
)
)
spec = torch.sqrt(spec.pow(2).sum(-1) + 1e-6)