§
    9²¸j:  ã                   óü   — d dl Z d dlmZ d dlZddlmZ ddlmZ ddl	m
Z
mZmZ  ej        e¦  «        Z	 ddeded	ed
ededededz  dej        fd„Zdej        dedededej        f
d„Z G d„ de¦  «        ZdgZdS )é    N)ÚSequenceé   )ÚSequenceFeatureExtractor)ÚBatchFeature)ÚPaddingStrategyÚ
TensorTypeÚloggingÚn_freqsÚf_minÚf_maxÚn_melsÚsample_rateÚ
fft_lengthÚnormÚreturnc                 ó2  — |�|dk    rt          d¦  «        ‚t          j        | t          j        ¬¦  «        ||z  z  }dt	          j        d|dz  z   ¦  «        z  }dt	          j        d|dz  z   ¦  «        z  }	t          j        ||	|dz   ¦  «        }
dd	|
dz  z  dz
  z  }|d
d…         |dd…         z
  }t          j        |d¦  «        t          j        |d
¦  «        z
  }t          j        d
t          j        ¬¦  «        }d|dd…dd…f         z  |dd…         z  }|dd…dd…f         |d
d…         z  }t          j	        |t          j
        ||¦  «        ¦  «        }|�9|dk    r3d|d|dz   …         |d|…         z
  z  }|t          j        |d¦  «        z  }|S )a©  Create a frequency bin conversion matrix (NumPy version).

    Args:
        n_freqs (int): Number of frequencies to highlight/apply
        f_min (float): Minimum frequency (Hz)
        f_max (float): Maximum frequency (Hz)
        n_mels (int): Number of mel filterbanks
        sample_rate (int): Sample rate of the audio waveform
        fft_length (int): FFT length
        norm (Optional[str]): If 'slaney', divide the triangular mel weights by
          the width of the mel band (area normalization). (Default: ``None``)

    Returns:
        np.ndarray: Triangular filter banks (fb matrix) of size (``n_freqs``,
        ``n_mels``)
        meaning number of frequencies to highlight/apply to x the number of
        filterbanks.
        Each column is a filterbank so that assuming there is a matrix A of
        size (..., ``n_freqs``), the applied result would be
        ``A @ create_fb_matrix_numpy(A.shape[-1], ...)``.
    NÚslaneyz$norm must be one of None or 'slaney'©Údtypeg     F¤@ç      ð?g     à…@é   é
   é   éÿÿÿÿr   g      ð¿éþÿÿÿg       @)Ú
ValueErrorÚnpÚarangeÚfloat32ÚmathÚlog10ÚlinspaceÚexpand_dimsÚzerosÚmaximumÚminimum)r
   r   r   r   r   r   r   Ú	all_freqsÚm_minÚm_maxÚm_ptsÚf_ptsÚf_diffÚslopesÚzeroÚdown_slopesÚ	up_slopesÚfbÚenorms                      ú†/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/transformers/models/gemma3n/feature_extraction_gemma3n.pyÚcreate_fb_matrixr4      s·  € ð> Ð˜D HÒ,Ð,ÝÐ?Ñ@Ô@Ð@õ ”	˜'­¬Ð4Ñ4Ô4¸ÀjÑ8PÑQ€Ið •T”Z  u¨u¡}Ñ 5Ñ6Ô6Ñ6€EØ•T”Z  u¨u¡}Ñ 5Ñ6Ô6Ñ6€EÝŒK˜˜u f¨q¡jÑ1Ô1€Eà�R˜E F™NÑ+¨cÑ1Ñ2€Eà�1�2�2ŒY˜˜s ˜sœÑ#€FÝŒ^˜E 1Ñ%Ô%­¬°yÀ!Ñ(DÔ(DÑD€FåŒ8�A�RœZÐ(Ñ(Ô(€DØ˜&    C R C œ.Ñ(¨F°3°B°3¬KÑ7€KØ�q�q�q˜!˜"˜"�u”  q r r¤
Ñ*€IÝ	Œ�D�"œ* [°)Ñ<Ô<Ñ	=Ô	=€BàÐ˜D HÒ,Ð,à�u˜Q ¨!¡˜^Ô,¨u°W°f°W¬~Ñ=Ñ>ˆØ
�bŒn˜U AÑ&Ô&Ñ&ˆà€Ió    ÚarrayÚ	dimensionÚsizeÚstepc                 ó˜  — | j         dk    rt          d¦  «        ‚|dk    r|| j         dz
  k    rt          d¦  «        ‚| j        \  }}||z
  |z  dz   }|dk    rt          j        |d|f| j        ¬¦  «        S |||f}| j        d         | j        d         |z  | j        d         f}t          j        j         	                    | ||¬¦  «        S )	zNA basic NumPy equivalent of PyTorch's unfold for 2D arrays along the last dim.r   zFThis unfold implementation currently supports 2D arrays (batch, time).r   r   zFThis unfold implementation only supports unfolding the last dimension.r   r   )ÚshapeÚstrides)
Úndimr   r;   r   r$   r   r<   ÚlibÚstride_tricksÚ
as_strided)	r6   r7   r8   r9   Ú
batch_sizeÚoriginal_lengthÚ
num_framesÚoutput_shapeÚoutput_stridess	            r3   Ú_unfoldrF   Y   sÝ   € à„z�Q‚€ÝÐaÑbÔbÐbØ�B‚€˜9¨¬
°Q©Ò6Ð6ÝÐaÑbÔbÐbà"'¤+Ñ€J�Ø! DÑ(¨TÑ1°AÑ5€Jà�Q‚€ÝŒx˜ Q¨Ð-°U´[ÐAÑAÔAÐAà 
¨DÐ1€LØ”m AÔ&¨¬°aÔ(8¸4Ñ(?ÀÄÈqÔAQÐR€NåŒ6Ô×*Ò*¨5¸ÈnÐ*Ñ]Ô]Ð]r5   c            "       ó´  ‡ — e Zd ZdZddgZ	 	 	 	 	 	 	 	 	 	 	 	 	 	 	 	 d.dededededededededededededededee         dz  dee         dz  f ˆ fd „Z	d!e
j        d"e
j        d#ee
j        e
j        f         fd$„Z	 	 	 	 	 	 d/d'e
j        ee         z  ee
j                 z  eee                  z  d(eez  ez  d)edz  d*ed+edz  d,eez  dz  dedz  d#efd-„Zˆ xZS )0ÚGemma3nAudioFeatureExtractoraT
  An audio feature extractor Universal Speech Models https://huggingface.co/papers/2303.01037.

    Args:
        feature_size (`int`, *optional*, defaults to 128):
            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).
        padding_value (`float`, *optional*, defaults to 0.0):
            Padding value used to pad the audio. Should correspond to silences.
        return_attention_mask (`bool`, *optional*, defaults to `True`):
            Whether to return the attention mask for the generated MEL spectrograms.
        frame_length_ms (`float`, *optional*, defaults to 32.0):
            The length of a frame in milliseconds.
        hop_length_ms (`float`, *optional*, defaults to 10.0):
            Length of the overlapping windows for the STFT used to obtain the Mel Frequency coefficients.
        min_frequency (`float`, *optional*, defaults to 125.0):
            The minimum frequency (in Hz) for the Mel filterbank.
        max_frequency (`float`, *optional*, defaults to 7600.0):
            The maximum frequency (in Hz) for the Mel filterbank.
        preemphasis (`float`, *optional*, defaults to 0.97):
            The preemphasis coefficient.
        preemphasis_htk_flavor (`bool`, *optional*, defaults to `True`):
            Whether to use HTK-style preemphasis.
        fft_overdrive (`bool`, *optional*, defaults to `True`):
            Whether to use FFT overdrive.
        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_scale_factor (`float`, *optional*, defaults to 1.0):
            Scaling factor applied to the input waveform.
        mel_floor (`float`, *optional*, defaults to 1e-05):
            Minimum value for Mel spectrograms to avoid log(0).
        per_bin_mean (`Optional[Sequence[float]]`, *optional*):
            Mean values for per-bin normalization.
        per_bin_stddev (`Optional[Sequence[float]]`, *optional*):
            Standard deviation values for per-bin normalization.
    Úinput_featuresÚinput_features_maské€   é€>  ç        Tç      @@ç      $@ç     @_@ç     °½@ç
×£p=
ï?r   çñhãˆµøä>NÚfeature_sizeÚsampling_rateÚpadding_valueÚreturn_attention_maskÚframe_length_msÚhop_length_msÚmin_frequencyÚmax_frequencyÚpreemphasisÚpreemphasis_htk_flavorÚfft_overdriveÚditherÚinput_scale_factorÚ	mel_floorÚper_bin_meanÚper_bin_stddevc           	      ó<  •—  t          ¦   «         j        d||||dœ|¤Ž || _        || _        |	| _        |
| _        || _        || _        || _        t          t          ||z  dz  ¦  «        ¦  «        | _        t          t          ||z  dz  ¦  «        ¦  «        | _        t          j        |t          j        ¬¦  «        | _        dt#          j        t#          j        | j        ¦  «        ¦  «        z  }| j        r|dz  }|| _        t          j        | j        t          j        ¬¦  «        }ddt          j        dt          j        z  |z  | j        z  ¦  «        z
  z  }|                     t          j        ¦  «        | _        t7          | j        dz  dz   |||| j        d |¬¦  «        | _        |�/t          j        |¦  «                             dd|¦  «        | _        nd | _        |�0t          j        |¦  «                             dd|¦  «        | _         d S d | _         d S )	N)rT   rU   rV   rW   g     @�@r   r   g      à?r   )r
   r   r   r   r   r   r   © )!ÚsuperÚ__init__rZ   r[   r\   r]   r^   r_   r`   ÚintÚroundÚframe_lengthÚ
hop_lengthr   r6   Úfloat64ra   r    ÚceilÚlog2r   r   r   ÚcosÚpiÚastypeÚwindowr4   rU   Úmel_filtersÚreshaperb   rc   )ÚselfrT   rU   rV   rW   rX   rY   rZ   r[   r\   r]   r^   r_   r`   ra   rb   rc   Úkwargsr   Úhann_arangerr   Ú	__class__s                        €r3   rg   z%Gemma3nAudioFeatureExtractor.__init__š   s  ø€ ð( 	�‰ŒÔð 	
Ø%Ø'Ø'Ø"7ð		
ð 	
ð
 ð	
ð 	
ð 	
ð +ˆÔØ*ˆÔØ&ˆÔØ&<ˆÔ#Ø*ˆÔØˆŒØ"4ˆÔÝ¥ m°oÑ&EÈÑ&NÑ OÔ OÑPÔPˆÔÝ�e M°MÑ$AÀFÑ$JÑKÔKÑLÔLˆŒÝœ )µ2´:Ð>Ñ>Ô>ˆŒà�$œ)¥D¤I¨dÔ.?Ñ$@Ô$@ÑAÔAÑAˆ
ØÔð 	Ø˜!‰OˆJØ$ˆŒå”i Ô 1½¼ÐDÑDÔDˆØ˜�BœF 1¥r¤u¡9¨{Ñ#:¸TÔ=NÑ#NÑOÔOÑOÑPˆØ—m’m¥B¤JÑ/Ô/ˆŒå+Ø”O qÑ(¨1Ñ,ØØØØÔ*ØØ!ð
ñ 
ô 
ˆÔð Ð#Ý "¤¨Ñ 6Ô 6× >Ò >¸qÀ!À\Ñ RÔ RˆDÔÐà $ˆDÔàÐ%Ý"$¤(¨>Ñ":Ô":×"BÒ"BÀ1ÀaÈÑ"VÔ"VˆDÔÐÐà"&ˆDÔÐÐr5   ÚwaveformÚattention_maskr   c                 ób  — |j         dk    rt          j        |d¬¦  «        }| j        dk    r;|| j        t          j        j        |j        Ž                      |j        ¦  «        z  z   }| j	        dk    r
|| j	        z  }| j
        dz   }t          |d|| j        ¬¦  «        }| j        dk    rz| j        rQ|dd	d…f         d| j        z
  z  }|ddd…f         | j        |dd	d
…f         z  z
  }t          j        ||gd¬¦  «        }n.|ddd	…f         | j        |dd	d…f         z  z
  }n|dd	d…f         }|| j        z  }t          j                             || j        d¬¦  «        }t          j        |¦  «        }	t          j        |	| j        ¦  «        }
t          j        t          j        |
| j        ¦  «        ¦  «        }| j        �
|| j        z
  }| j        �
|| j        z  }|                     d¦  «        }|d	d	| j        …                              t:          ¦  «        }||d	|j        d         …         fS )Ú r   r   )ÚaxisrM   r   r   )r7   r8   r9   .Nr   )Únr}   )r=   r   r#   r_   ÚrandomÚrandnr;   rq   r   r`   rj   rF   rk   r\   r]   Úconcatenaterr   ÚfftÚrfftr   ÚabsÚmatmulrs   Úlogr%   ra   rb   rc   ÚsqueezeÚbool)ru   ry   rz   Úframe_size_for_unfoldÚframes_to_processÚfirst_in_frameÚrest_in_frameÚframesÚstftÚmagnitude_specÚmel_specÚlog_mel_specÚmel_spectrogramÚmasks                 r3   Ú_extract_spectrogramz1Gemma3nAudioFeatureExtractor._extract_spectrogramÞ   sV  € àŒ=˜AÒÐÝ”~ h°QÐ7Ñ7Ô7ˆHàŒ;˜ÒÐØ $¤+µ´	´ÀÄÐ0P×0WÒ0WÐX`ÔXfÑ0gÔ0gÑ"gÑgˆHàÔ" cÒ)Ð)Ø $Ô"9Ñ9ˆHà $Ô 1°AÑ 5Ðõ $ H¸ÐAVÐ]aÔ]lÐmÑmÔmÐàÔ˜cÒ!Ð!ØÔ*ð eØ!2°3¸¸¸°7Ô!;¸sÀTÔEUÑ?UÑ!V�Ø 1°#°q¸°t°)Ô <¸tÔ?OÐRcÐdgÐilÐjlÐilÐdlÔRmÑ?mÑ m�Ýœ¨¸Ð(GÈbÐQÑQÔQ��à*¨3°°°¨7Ô3°dÔ6FÐIZÐ[^Ð`cÐacÐ`cÐ[cÔIdÑ6dÑd��à& s¨C¨R¨C xÔ0ˆFà˜$œ+Ñ%ˆÝŒv�{Š{˜6 T¤_¸2ˆ{Ñ>Ô>ˆåœ ™œˆå”9˜^¨TÔ-=Ñ>Ô>ˆÝ”v�bœj¨°4´>ÑBÔBÑCÔCˆàÔÐ(Ø'¨$Ô*;Ñ;ˆLàÔÐ*Ø'¨$Ô*=Ñ=ˆLà&×.Ò.¨qÑ1Ô1ˆØÐ0Ð0 ¤Ð0Ô1×8Ò8½Ñ>Ô>ˆà Ð%? Ô'<¸QÔ'?Ð%?Ô @Ð@Ð@r5   Úlongesté S Ú
raw_speechÚpaddingÚ
max_lengthÚ
truncationÚpad_to_multiple_ofÚreturn_tensorsc                 ó–  — t          |t          j        ¦  «        ot          |j        ¦  «        dk    }	t          |t
          ¦  «        o&t          |d         t          j        t
          f¦  «        }
|	p|
}|s|g}d„ |D ¦   «         }|                      t          d|i¦  «        |||||¬¦  «        }g }g }t          |j	        |j
        ¦  «        D ]j\  }}|                      |j        |¦  «        \  }}|                     |                     t          j        ¦  «        ¦  «         |                     |¦  «         Œkt          ||dœ|¬¦  «        S )a  Creates a batch of MEL spectrograms from the provided raw speech.

        This implementation uses a different algorithm for windowing and preemphasis compared to the built-in
        `transformers.audio_utils.spectrogram()` function that _will_ result in different outputs. Consider this
        carefully when selecting an audio feature extractor, especially with pre-trained models.

        Args:
            raw_speech:
                The audio for which MEL spectrograms are created.
            padding (`Union[bool, str, PaddingStrategy]`, *optional*, defaults to `"longest"`):
                The padding strategy to use for batches of audio with different lengths.
            max_length (`int`, *optional*, defaults to 480000):
                If provided, defines the maximum length of the audio to allow. Audio longer than this will be
                truncated if `truncation=True`.
            truncation (`bool`, *optional*, defaults to `True`):
                Whether or not to truncate audio above `max_length`.
            pad_to_multiple_of (`int`, *optional*, defaults to 128):
                When padding, pad to a multiple of this value. The default value is defined for optimal TPU support.
            return_tensors (`Union[str, TensorType]`, *optional*, defaults to `None`):
                The type of tensors to return (e.g., NumPy, or Torch).
            return_attention_mask (`bool`, *optional*, defaults to `True`):
                Whether to return the attention mask for the generated MEL spectrograms.
        r   r   c                 óB   — g | ]}t          j        |g¦  «        j        ‘ŒS re   )r   ÚasarrayÚT)Ú.0Úrss     r3   ú
<listcomp>z9Gemma3nAudioFeatureExtractor.__call__.<locals>.<listcomp>5  s'   € Ð>Ð>Ð>¨R•b”j " Ñ&Ô&Ô(Ð>Ð>Ð>r5   rI   )r˜   r™   rš   r›   rW   )rI   rJ   )Útensor_type)Ú
isinstancer   ÚndarrayÚlenr;   r   Úpadr   ÚziprI   rz   r”   r    Úappendrq   r   )ru   r—   r˜   r™   rš   r›   rœ   rW   rv   Úis_batched_numpyÚis_batched_sequenceÚ
is_batchedÚbatched_speechÚprepared_speechÚprepared_speech_maskÚspeechr“   s                    r3   Ú__call__z%Gemma3nAudioFeatureExtractor.__call__  sl  € õF & jµ"´*Ñ=Ô=Ð[Å#ÀjÔFVÑBWÔBWÐZ[ÒB[ÐÝ(¨µXÑ>Ô>ÐtÅ:ÈjÐYZÌmÕ^`Ô^hÕjrÐ]sÑCtÔCtÐØ%Ð<Ð)<ˆ
ð ð 	&Ø$˜ˆJØ>Ð>°:Ð>Ñ>Ô>ˆ
àŸšÝÐ*¨JÐ7Ñ8Ô8ØØ!Ø!Ø1Ø"7ð "ñ 
ô 
ˆð ˆØ!ÐÝ Ô =¸~Ô?\Ñ]Ô]ð 	.ð 	.‰LˆF�DØ×4Ò4°V´X¸tÑDÔD‰LˆF�DØ×"Ò" 6§=¢=µ´Ñ#<Ô#<Ñ=Ô=Ð=Ø ×'Ò'¨Ñ-Ô-Ð-Ð-åØ.ÐG[Ð\Ð\Ø&ð
ñ 
ô 
ð 	
r5   )rK   rL   rM   TrN   rO   rP   rQ   rR   TTrM   r   rS   NN)r•   r–   TrK   NT)Ú__name__Ú
__module__Ú__qualname__Ú__doc__Úmodel_input_namesrh   Úfloatrˆ   r   rg   r   r¦   Útupler”   ÚlistÚstrr   r   r   r²   Ú__classcell__)rx   s   @r3   rH   rH   l   s\  ø€ € € € € ð)ð )ðV *Ð+@ÐAÐð  Ø#Ø"Ø&*Ø!%Ø#Ø$Ø%Ø!Ø'+Ø"ØØ$'ØØ/3Ø15ð#B'ð B'àðB'ð ðB'ð ð	B'ð
  $ðB'ð ðB'ð ðB'ð ðB'ð ðB'ð ðB'ð !%ðB'ð ðB'ð ðB'ð "ðB'ð ðB'ð  ˜u”o¨Ñ,ð!B'ð" ! œ¨$Ñ.ð#B'ð B'ð B'ð B'ð B'ð B'ðH+A¨R¬Zð +AÈÌð +AÐX]Ð^`Ô^hÐjlÔjtÐ^tÔXuð +Að +Að +Að +Að` 1:Ø!(ØØ),Ø26Ø-1ð?
ð ?
à”J  e¤Ñ,¨t°B´JÔ/?Ñ?À$ÀtÈEÄ{ÔBSÑSð?
ð ˜‘˜oÑ-ð?
ð ˜$‘Jð	?
ð
 ð?
ð   $™Jð?
ð ˜jÑ(¨4Ñ/ð?
ð  $ d™{ð?
ð 
ð?
ð ?
ð ?
ð ?
ð ?
ð ?
ð ?
ð ?
r5   rH   )N)r    Úcollections.abcr   Únumpyr   Ú!feature_extraction_sequence_utilsr   Úfeature_extraction_utilsr   Úutilsr   r   r	   Ú
get_loggerr³   Úloggerrh   r¸   r»   r¦   r4   rF   rH   Ú__all__re   r5   r3   ú<module>rÅ      s{  ðð €€€Ø $Ð $Ð $Ð $Ð $Ð $à Ð Ð Ð à IÐ IÐ IÐ IÐ IÐ IØ 4Ð 4Ð 4Ð 4Ð 4Ð 4Ø 9Ð 9Ð 9Ð 9Ð 9Ð 9Ð 9Ð 9Ð 9Ð 9ð 
ˆÔ	˜HÑ	%Ô	%€ð ð:ð :Øð:àð:ð ð:ð ð	:ð
 ð:ð ð:ð �‰*ð:ð „Zð:ð :ð :ð :ðz^�2”:ð ^¨#ð ^°Sð ^Àð ^ÈÌ
ð ^ð ^ð ^ð ^ð&^
ð ^
ð ^
ð ^
ð ^
Ð#;ñ ^
ô ^
ð ^
ðB *Ð
*€€€r5   