deepgeek/CosyVoice
2
0
Fork 0
mirror of https://github.com/FunAudioLLM/CosyVoice.git synced 2026-02-04 09:29:25 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix lint

Browse Source
This commit is contained in:
lyuxiang.lx
2025-02-06 16:07:13 +08:00
parent 24f796a2b1
commit 2a3e033ee1
17 changed files with 187 additions and 135 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
cosyvoice/dataset/processor.py
View File

@@ -196,8 +196,8 @@ def compute_f0(data, sample_rate, hop_size, mode='train'):
assert 'text_token' in sample
waveform = sample['speech']
_f0, t = pw.harvest(waveform.squeeze(dim=0).numpy().astype('double'), sample_rate, frame_period=frame_period)
if sum(_f0 != 0) < 5: # this happens when the algorithm fails
_f0, t = pw.dio(waveform.squeeze(dim=0).numpy().astype('double'), sample_rate, frame_period=frame_period) # if harvest fails, try dio
if sum(_f0 != 0) < 5: # this happens when the algorithm fails
_f0, t = pw.dio(waveform.squeeze(dim=0).numpy().astype('double'), sample_rate, frame_period=frame_period) # if harvest fails, try dio
f0 = pw.stonemask(waveform.squeeze(dim=0).numpy().astype('double'), _f0, t, sample_rate)
f0 = F.interpolate(torch.from_numpy(f0).view(1, 1, -1), size=sample['speech_feat'].shape[0], mode='linear').view(-1)
sample['pitch_feat'] = f0