heimoshuiyu/whisper
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Support batch-dimension in log_mel_spectogram (#839)

Browse Source
This commit is contained in:
Markus Hennerbichler
2023-01-17 07:46:15 +00:00
committed by GitHub
parent 70861c7ce3
commit 6df3ea1fb5
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
whisper/audio.py
View File

@@ -113,7 +113,7 @@ def log_mel_spectrogram(audio: Union[str, np.ndarray, torch.Tensor], n_mels: int
window = torch.hann_window(N_FFT).to(audio.device) window = torch.hann_window(N_FFT).to(audio.device)
stft = torch.stft(audio, N_FFT, HOP_LENGTH, window=window, return_complex=True) stft = torch.stft(audio, N_FFT, HOP_LENGTH, window=window, return_complex=True)
magnitudes = stft[:, :-1].abs() ** 2 magnitudes = stft[..., :-1].abs() ** 2
filters = mel_filters(audio.device, n_mels) filters = mel_filters(audio.device, n_mels)
mel_spec = filters @ magnitudes mel_spec = filters @ magnitudes