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

.

Browse Source
This commit is contained in:
Mahmoud Ashraf
2023-10-16 20:43:37 +03:00
committed by GitHub
parent 02c0323777
commit b69956d725
1 changed files with 9 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
whisperx/alignment.py
View File

@ -207,17 +207,17 @@ def align(
# TODO: Probably can get some speedup gain with batched inference here
waveform_segment = audio[:, f1:f2]
# Handle the minimum input length for wav2vec2 models
if waveform_segment.shape[-1] < 400:
lengths = torch.as_tensor([waveform_segment.shape[-1]]).to(device)
waveform_segment = torch.nn.functional.pad(
waveform_segment, (0, 400 - waveform_segment.shape[-1])
)
else:
lengths = None
with torch.inference_mode():
if model_type == "torchaudio":
# Handle the minimum input length for torchaudio wav2vec2 models
if waveform_segment.shape[-1] < 400:
lengths = torch.as_tensor([waveform_segment.shape[-1]]).to(device)
waveform_segment = torch.nn.functional.pad(
waveform_segment, (0, 400 - waveform_segment.shape[-1])
)
else:
lengths = None
emissions, _ = model(waveform_segment.to(device), lengths=lengths)
elif model_type == "huggingface":
emissions = model(waveform_segment.to(device)).logits