o 9wi> @s.ddlZddlZddlmZddlmZmZddlZddl Z ddl m m Z ddlmZdZdZdZdZeeZeeeZed ZeeeZeeeZefd ed ed ejfd dZefdddedefddZeddded e jfddZ ddeeeje jfdededeeee j ffddZ!dS)N) lru_cache)OptionalUnion) exact_divi>ifilesrreturncCszddddd|dddd d d d t|d g}tj|dddj}Wntjy6}z td|j|d}~wwt |tj  tj dS)a? Open an audio file and read as mono waveform, resampling as necessary Parameters ---------- file: str The audio file to open sr: int The sample rate to resample the audio if necessary Returns ------- A NumPy array containing the audio waveform, in float32 dtype. ffmpegz-nostdinz-threads0z-iz-fs16lez-ac1z-acodec pcm_s16lez-ar-T)capture_outputcheckzFailed to load audio: Ng@)str subprocessrunstdoutCalledProcessError RuntimeErrorstderrdecodenp frombufferint16flattenastypefloat32)r r cmdouter%K/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/whisperx/audio.py load_audios.r')axislengthr)cCst|rC|j||kr|j|tj||jdd}|j||krAdg|j}d||j|f||<t|dd|dddD}|S|j||krS|j t ||d }|j||krqdg|j}d||j|f||<t ||}|S) zO Pad or trim the audio array to N_SAMPLES, as expected by the encoder. )device)dimindex)rrrcSsg|] }|D]}|qqSr%r%).0sizespadr%r%r& Qszpad_or_trim..Nr()indicesr)) torch is_tensorshape index_selectaranger+ndimFr0takeranger)arrayr*r) pad_widthsr%r%r& pad_or_trimDs"      r>)maxsizen_melscCsr|dvs Jd|ttjtjtdd}t|d| |WdS1s2wYdS)a load the mel filterbank matrix for projecting STFT into a Mel spectrogram. Allows decoupling librosa dependency; saved using: np.savez_compressed( "mel_filters.npz", mel_80=librosa.filters.mel(sr=16000, n_fft=400, n_mels=80), ) )PzUnsupported n_mels: assetszmel_filters.npzmel_N) rloadospathjoindirname__file__r3 from_numpyto)r+r@fr%r%r& mel_filters^s $rNaudiopaddingr+c Cst|st|trt|}t|}|dur||}|dkr(t|d|f}t t |j }tj |t t |dd}|dddfd}t|j |}||}tj|dd } t| | d } | d d } | S) ap Compute the log-Mel spectrogram of Parameters ---------- audio: Union[str, np.ndarray, torch.Tensor], shape = (*) The path to audio or either a NumPy array or Tensor containing the audio waveform in 16 kHz n_mels: int The number of Mel-frequency filters, only 80 is supported padding: int Number of zero samples to pad to the right device: Optional[Union[str, torch.device]] If given, the audio tensor is moved to this device before STFT Returns ------- torch.Tensor, shape = (80, n_frames) A Tensor that contains the Mel spectrogram NrT)windowreturn_complex.r(rg|=)ming @g@)r3r4 isinstancerr'rKrLr9r0 hann_windowN_FFTr+stft HOP_LENGTHabsrNclamplog10maximummax) rOr@rPr+rQrW magnitudesfiltersmel_speclog_specr%r%r&log_mel_spectrogramps"      rb)rN)"rFr functoolsrtypingrrnumpyrr3torch.nn.functionalnn functionalr9whisperx.utilsr SAMPLE_RATErVrX CHUNK_LENGTH N_SAMPLESN_FRAMESN_SAMPLES_PER_TOKENFRAMES_PER_SECONDTOKENS_PER_SECONDrintndarrayr'r>TensorrNr+rbr%r%r%r&s@     +