o f2hD @s2ddlZddlmZddlmZmZddlmZmZddl Z ddl Z ddl m mZddlmZdZdZd Zd ZeeZeeeZed ZeeeZeeeZefd ed efddZefdddedefddZedddede jfddZ  d!deee j!e jfdededeeee j"ffdd Z#dS)"N) lru_cache)CalledProcessErrorrun)OptionalUnion) exact_divi>ifilesrcCsddddd|dddd d d d t|d g}z t|dddj}Wnty4}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@) strrstdoutr RuntimeErrorstderrdecodenp frombufferint16flattenastypefloat32)r r cmdouter$a/var/www/html/alexa/alex_system/speach-to-text/venv/lib/python3.10/site-packages/whisper/audio.py load_audios" r&)axislengthr(cCst|rC|j||kr|j|tj||jdd}|j||krAdg|j}d||j|f||<t|dd|dddD}|S|j||krS|j t ||d }|j||krqdg|j}d||j|f||<t ||}|S) zO Pad or trim the audio array to N_SAMPLES, as expected by the encoder. )device)dimindex)rrrcSsg|] }|D]}|qqSr$r$).0sizespadr$r$r% Nszpad_or_trim..Nr')indicesr() torch is_tensorshape index_selectaranger*ndimFr/takeranger)arrayr)r( pad_widthsr$r$r% pad_or_trimAs"      r=)maxsizen_melsreturncCsz|dvs Jd|tjtjtdd}tj|dd}t|d| |WdS1s6wYdS) ad load the mel filterbank matrix for projecting STFT into a Mel spectrogram. Allows decoupling librosa dependency; saved using: np.savez_compressed( "mel_filters.npz", mel_80=librosa.filters.mel(sr=16000, n_fft=400, n_mels=80), mel_128=librosa.filters.mel(sr=16000, n_fft=400, n_mels=128), ) >PzUnsupported n_mels: assetszmel_filters.npzF) allow_picklemel_N) ospathjoindirname__file__rloadr2 from_numpyto)r*r? filters_pathfr$r$r% mel_filters[s  $rPrAaudiopaddingr*c Cst|st|trt|}t|}|dur||}|dkr(t|d|f}t t |j }tj |t t |dd}|dddfd}t|j |}||}tj|dd } t| | d } | d d } | S) ap Compute the log-Mel spectrogram of Parameters ---------- audio: Union[str, np.ndarray, torch.Tensor], shape = (*) The path to audio or either a NumPy array or Tensor containing the audio waveform in 16 kHz n_mels: int The number of Mel-frequency filters, only 80 is supported padding: int Number of zero samples to pad to the right device: Optional[Union[str, torch.device]] If given, the audio tensor is moved to this device before STFT Returns ------- torch.Tensor, shape = (80, n_frames) A Tensor that contains the Mel spectrogram NrT)windowreturn_complex.r'r g|=)ming @g@)r2r3 isinstancerr&rLrMr8r/ hann_windowN_FFTr*stft HOP_LENGTHabsrPclamplog10maximummax) rQr?rRr*rSrY magnitudesfiltersmel_speclog_specr$r$r%log_mel_spectrogramns"      rd)rArN)$rF functoolsr subprocessrrtypingrrnumpyrr2torch.nn.functionalnn functionalr8utilsr SAMPLE_RATErXrZ CHUNK_LENGTH N_SAMPLESN_FRAMESN_SAMPLES_PER_TOKENFRAMES_PER_SECONDTOKENS_PER_SECONDrintr&r=TensorrPndarrayr*rdr$r$r$r%sB     (