chrissewell/whisperX
1
0
Fork 0
mirror of https://github.com/m-bain/whisperX.git synced 2025-07-01 18:17:27 -04:00
Code Issues Packages Projects Releases Wiki Activity

fix errors when recognizing non-japanese audios

Browse Source
This commit is contained in:
Yasutaka Odo
2022-12-21 01:42:45 +09:00
parent 7a3575dfdf
commit d7546def91
1 changed files with 5 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
whisperx/transcribe.py
View File

@ -282,8 +282,12 @@ def align(
f2 = int(t2 * SAMPLE_RATE) f2 = int(t2 * SAMPLE_RATE)
waveform_segment = audio[:, f1:f2] waveform_segment = audio[:, f1:f2]
print(language)
with torch.inference_mode(): with torch.inference_mode():
emissions = model(waveform_segment.to(device)).logits if language != 'ja':
emissions, _ = model(waveform_segment.to(device))
else:
emissions = model(waveform_segment.to(device)).logits
emissions = torch.log_softmax(emissions, dim=-1) emissions = torch.log_softmax(emissions, dim=-1)
emission = emissions[0].cpu().detach() emission = emissions[0].cpu().detach()
transcription = segment['text'].strip() transcription = segment['text'].strip()