o Zh=@sdZddlmZmZmZddlZddlmZddl m Z m Z m Z ddl mZddlmZdd lmZmZer;ddlZeeZGd d d eZd gZdS) z% Feature extractor class for Whisper )ListOptionalUnionN)is_torch_available)mel_filter_bank spectrogramwindow_function)SequenceFeatureExtractor) BatchFeature) TensorTypeloggingcs>eZdZdZdgZ        d(fd d Zd ejd edej fddZ d)dejd edej fddZ e d*de ej de ej dede ej fddZ          d+deej e ee ej e e efdedeed eeeefd!eed"eedeed#eed$eed eed%eedefd&d'ZZS),WhisperFeatureExtractora Constructs a Whisper feature extractor. This feature extractor inherits from [`~feature_extraction_sequence_utils.SequenceFeatureExtractor`] which contains most of the main methods. Users should refer to this superclass for more information regarding those methods. This class extracts mel-filter bank features from raw speech using a custom numpy implementation of the `Short Time Fourier Transform` which should match pytorch's `torch.stft` equivalent. Args: feature_size (`int`, *optional*, defaults to 80): The feature dimension of the extracted features. sampling_rate (`int`, *optional*, defaults to 16000): The sampling rate at which the audio files should be digitalized expressed in hertz (Hz). hop_length (`int`, *optional*, defaults to 160): Length of the overlapping windows for the STFT used to obtain the Mel Frequency coefficients. chunk_length (`int`, *optional*, defaults to 30): The maximum number of chunks of `sampling_rate` samples used to trim and pad longer or shorter audio sequences. n_fft (`int`, *optional*, defaults to 400): Size of the Fourier transform. padding_value (`float`, *optional*, defaults to 0.0): Padding value used to pad the audio. Should correspond to silences. dither (`float`, *optional*, defaults to 0.0): Adds dithering. In other words, adds a small Gaussian noise to each frame. E.g. use 0.0001 to add dithering with a normal distribution centered around 0.0 with standard deviation 0.0001 (assuming [-1,+1] range of raw_speech). The value 0.0 means no dithering. Dithering has similar effect as `spectrogram(mel_floor=...)`. It reduces the high log_mel_fbank values for signals with hard-zero sections, when VAD cutoff is present in the signal. input_featuresP>Fc sttjd||||d| ||_||_||_|||_|j||_||_||_t d|d|dd|ddd|_ dS) N) feature_size sampling_rate padding_valuereturn_attention_maskrg@@Zslaney)Znum_frequency_binsZnum_mel_filtersZ min_frequencyZ max_frequencyrZnormZ mel_scale) super__init__n_fft hop_length chunk_length n_samplesZ nb_max_framesrditherr mel_filters) selfrrr r!rrr#rkwargs __class__re/var/www/auris/lib/python3.10/site-packages/transformers/models/whisper/feature_extraction_whisper.pyrHs0     z WhisperFeatureExtractor.__init__waveform_batchdevicereturnc Cs|dkr td|dg}|D]5}t|t|jd|j|jd|j|jdd}|dddd f}t|| d }|d d }| |qt |}|S) z Compute the log-mel spectrogram of the provided audio, gives similar results to Whisper's original torch implementation with 1e-5 tolerance. cpuz Got device `z` for feature extraction, but feature extraction on CUDA accelerator devices requires torch, which is not installed. Either set `device='cpu'`, or install torch according to the official instructions: https://pytorch.org/get-started/locally/Zhanng@log10)Z frame_lengthr powerr#r$Zlog_melN @@) ValueErrorrr rr r#r$npmaximummaxappendarray)r%r*r+Zlog_spec_batchwaveformlog_specrrr)_np_extract_fbank_featuresls,      z2WhisperFeatureExtractor._np_extract_fbank_featuresr-r9c Cs*t||tj}tj|j|d}|jdkr'||jtj|j|j |j d7}tj ||j|j |dd}|dddf d }t|j|tj}|j|}tj|d d }|d krw|jd dd d jddd d } t|| d}n t||d}|dd}|dkr|}|S)z Compute the log-mel spectrogram of the audio using PyTorch's GPU-accelerated STFT implementation with batching, yielding results similar to cpu computing with 1e-5 tolerance. )r+r)dtyper+T)windowZreturn_complex.Nr0rg|=)min)dimZkeepdimrrr1r2r-)torchZ from_numpytofloat32Z hann_windowrr#Zrandnshaper<r+stftr absr$Tclampr.r?r6r5detachr-numpy) r%r9r+r=rDZ magnitudesr$Zmel_specr:Zmax_valrrr)_torch_extract_fbank_featuress"       z5WhisperFeatureExtractor._torch_extract_fbank_features input_valuesattention_maskrcCs|durEt|tj}g}t||dD]-\}}||d|t|d|d}||jdkr=|||d<| |q|Sdd|D}|S)z[ Every array in the list is normalized to have zero mean and unit variance Nr0Hz>rcSs*g|]}||t|dqS)rM)meanr4sqrtvar).0xrrr) s*zCWhisperFeatureExtractor.zero_mean_unit_var_norm..) r4r8Zint32zipsumrNrOrPrCr7)rKrLrZnormed_input_valuesZvectorlengthZ normed_slicerrr)zero_mean_unit_var_norms.  z/WhisperFeatureExtractor.zero_mean_unit_var_normTN max_length raw_speech truncationpad_to_multiple_ofreturn_tensorsrpaddingr do_normalizereturn_token_timestampsc  sN|dur|jkrtdjjdjdjd|d n tdjjdt|tjo6t |j d k} | rGt |j d krGtd | pZt|t t foZt|d tjt t f}|red d|D}n&|svt|tjsvtj |tjd}nt|tjr|jttjur|tj}|st |gjg}td|i}j|||r|nj|||p| d}| rɈj|d|djd|d<tj|dd d|d<|dd d d }trوjnj}||d | }t|d trdd|D|d<n||d<|r |dddddjf|d<| durfdd|D|d<|dur%| |}|S)a Main method to featurize and prepare for the model one or several sequence(s). Implementation uses PyTorch for the STFT computation if available, otherwise a slower NumPy based one. Args: raw_speech (`np.ndarray`, `List[float]`, `List[np.ndarray]`, `List[List[float]]`): The sequence or batch of sequences to be padded. Each sequence can be a numpy array, a list of float values, a list of numpy arrays or a list of list of float values. Must be mono channel audio, not stereo, i.e. single float per timestep. truncation (`bool`, *optional*, default to `True`): Activates truncation to cut input sequences longer than *max_length* to *max_length*. pad_to_multiple_of (`int`, *optional*, defaults to None): If set will pad the sequence to a multiple of the provided value. This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability `>= 7.5` (Volta), or on TPUs which benefit from having sequence lengths be a multiple of 128. return_attention_mask (`bool`, *optional*): Whether to return the attention mask. If left to the default, will return the attention mask according to the specific feature_extractor's default. [What are attention masks?](../glossary#attention-mask) For Whisper models, `attention_mask` should always be passed for batched inference, to avoid subtle bugs. return_tensors (`str` or [`~utils.TensorType`], *optional*): If set, will return tensors instead of list of python integers. Acceptable values are: - `'tf'`: Return TensorFlow `tf.constant` objects. - `'pt'`: Return PyTorch `torch.Tensor` objects. - `'np'`: Return Numpy `np.ndarray` objects. sampling_rate (`int`, *optional*): The sampling rate at which the `raw_speech` input was sampled. It is strongly recommended to pass `sampling_rate` at the forward call to prevent silent errors and allow automatic speech recognition pipeline. padding_value (`float`, *optional*, defaults to 0.0): The value that is used to fill the padding values / vectors. do_normalize (`bool`, *optional*, defaults to `False`): Whether or not to zero-mean unit-variance normalize the input. Normalizing can help to significantly improve the performance of the model. device (`str`, *optional*, defaults to `'cpu'`): Specifies the device for computation of the log-mel spectrogram of audio signals in the `_torch_extract_fbank_features` method. (e.g., "cpu", "cuda") return_token_timestamps (`bool`, *optional*, defaults to `None`): Whether or not to return the number of frames of the input raw_speech. These num_frames can be used by the model to compute word level timestamps. Nz3The model corresponding to this feature extractor: z& was trained using a sampling rate of zI. Please make sure that the provided `raw_speech` input was sampled with z and not .zDIt is strongly recommended to pass the `sampling_rate` argument to `zN()`. Failing to do so can result in silent errors that might be hard to debug.rrz2Only mono-channel audio is supported for input to rcSs g|] }tj|gtjdjqSr<)r4asarrayrBrF)rQZspeechrrr)rSs z4WhisperFeatureExtractor.__call__..rbr)r]rXrZr[rrL)rLr)ZaxiscSsg|] }tj|tjdqSra)r4rcrB)rQfeaturerrr)rS@scsg|] }t|jqSr)lenr )rQZ raw_speech_ir%rr)rSJsZ num_frames)!rr3r(__name__loggerwarning isinstancer4ndarrayrerClisttuplercrBr<Zfloat64ZastyperFr padr"rWrstackgetZ transposerrJr;rr Zconvert_to_tensors)r%rYrZr[r\rr]rXrr^r+r_r&Zis_batched_numpyZ is_batchedZbatched_speechZ padded_inputsrZextract_fbank_featuresrrfr)__call__svB  "         z WhisperFeatureExtractor.__call__)rrrrrrrF)r-)r) TNNNrXNNNr-N)rg __module__ __qualname____doc__Zmodel_input_namesrr4r8strrkr;rJ staticmethodrfloatrWrboolrintr r rq __classcell__rrr'r)r$sx!$ "    r)rttypingrrrrIr4rZ audio_utilsrrr Z!feature_extraction_sequence_utilsr Zfeature_extraction_utilsr utilsr r r@Z get_loggerrgrhr__all__rrrr)s     0