azalea/VITS-fast-fine-tuning
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

upload files

Browse Source
This commit is contained in:
Plachta
2023-02-13 17:20:19 +08:00
parent ea90485218
commit 1127f111d3
7 changed files with 756 additions and 410 deletions
Download Patch File Download Diff File Expand all files Collapse all files
data_utils.py
+23 -16
View File
@@ -4,6 +4,7 @@ import random
import numpy as np
import torch
import torch.utils.data
import torchaudio
import commons
from mel_processing import spectrogram_torch
@@ -190,7 +191,8 @@ class TextAudioSpeakerLoader(torch.utils.data.Dataset):
audiopaths_sid_text_new = []
lengths = []
for audiopath, sid, text in self.audiopaths_sid_text:
audiopath = "E:/uma_voice/" + audiopath
# audiopath = "./user_voice/" + audiopath
if self.min_text_len <= len(text) and len(text) <= self.max_text_len:
audiopaths_sid_text_new.append([audiopath, sid, text])
lengths.append(os.path.getsize(audiopath) // (2 * self.hop_length))
@@ -206,21 +208,26 @@ class TextAudioSpeakerLoader(torch.utils.data.Dataset):
return (text, spec, wav, sid)
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)
# 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 if audio.max() > 10 else audio
# audio_norm = audio_norm.unsqueeze(0)
audio_norm, sampling_rate = torchaudio.load(filename, frame_offset=0, num_frames=-1, normalize=True, channels_first=True)
# spec_filename = filename.replace(".wav", ".spec.pt")
# if os.path.exists(spec_filename):
# spec = torch.load(spec_filename)
# else:
# try:
spec = spectrogram_torch(audio_norm, self.filter_length,
self.sampling_rate, self.hop_length, self.win_length,
center=False)
spec = spec.squeeze(0)
# except NotImplementedError:
# print("?")
# spec = torch.squeeze(spec, 0)
# torch.save(spec, spec_filename)
return spec, audio_norm
def get_text(self, text):