+ j:^RIt^RIHt^RIt^RIHt^RIHt^RI H t H t H t ] P!]4tR RRlltRR lt!R R ]4tR .tR#) N)Sequence)SequenceFeatureExtractor) BatchFeature)PaddingStrategy TensorTypeloggingcV^8dQhR\R\R\R\R\R\R\R,R \P/#) n_freqsf_minf_maxn_mels sample_rate fft_lengthnormNreturn)intfloatstrnpndarray)formats"w/home/luca/screen-checking/.venv/lib/python3.14/site-packages/transformers/models/gemma3n/feature_extraction_gemma3n.py __annotate__rs`:: : : :  :  :  : *:ZZ:cVeVR8wd \R4h\P!V\PR7WE, ,pR\P !RVR, ,4,pR\P !RVR, ,4,p \P !WV^,4p R^ V R, ,R, ,p V R,V RR , p \P!V ^4\P!V^4, p \P!^\PR7pR V R RR 13,,V RR , pV R,V R,, p\P!V\P!VV44pVeCVR8Xd<R V ^V^,V RV, , pV\P!V^4,pV#)aaCreate 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], ...)``. Nslaneyz$norm must be one of None or 'slaney'dtypegF@?g@NNNNNg@g)r#:r NN) ValueErrorrarangefloat32mathlog10linspace expand_dimszerosmaximumminimum)r r r rrrr all_freqsm_minm_maxm_ptsf_ptsf_diffslopeszero down_slopes up_slopesfbenorms&&&&&&& rcreate_fb_matrixr<s~> DH,?@@ '4 8PQI TZZuu} 56 6E TZZuu} 56 6E KKfqj 1E REFN+c1 2E 2Ys #F ^^E1 %y!(D DF 88ARZZ (D&CRC.(F3BK7Ku r *I D"**[)< =B DH,uQ!,uWf~=> bnnUA&& Irc |V^8dQhR\PR\R\R\R\P/#)r array dimensionsizestepr)rrr)rs"rrrYs>^^2::^#^S^^ ^rcVP^8wd \R4hVR8wd#WP^, 8wd \R4hVPwrEWR, V,^,pV^8:d&\P!V^V3VP R7#WFV3pVP ^,VP ^,V,VP ^,3p\PPPWVR7#)zNA basic NumPy equivalent of PyTorch's unfold for 2D arrays along the last dim.zFThis unfold implementation currently supports 2D arrays (batch, time).zFThis unfold implementation only supports unfolding the last dimension.r)shapestridesr$) ndimr&rCrr-rrDlib stride_tricks as_strided) r>r?r@rA batch_sizeoriginal_length num_frames output_shapeoutput_stridess &&&& r_unfoldrNYs zzQabbB9 Q6abb"'++J!(T1A5JQxxQ-U[[AAD1LmmA& a(84(?qAQRN 66   * *5n * ]]rcvaa]tRt^ltoRtRR.tR V3RlV3RllltV3RlRltR V3RlR lltR t Vt V;t #) Gemma3nAudioFeatureExtractora 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. input_featuresinput_features_maskc!<V^8dQhRS[RS[RS[RS[RS[RS[RS[RS[R S[R S[R S[R S[R S[RS[RS[S[,R,RS[S[,R,/#)r feature_size sampling_rate padding_valuereturn_attention_maskframe_length_ms hop_length_ms min_frequency max_frequency preemphasispreemphasis_htk_flavor fft_overdriveditherinput_scale_factor mel_floor per_bin_meanNper_bin_stddev)rrboolr)r __classdict__s"rr)Gemma3nAudioFeatureExtractor.__annotate__sB'B'B'B' B' $ B'  B'B'B'B'B'!%B'B'B'"B'B' uo,!B'"!$.#B'rc <\SV`!R RVRVRVRV/VBWpnWnWnWnWnWnWn\\W%,R, 44Vn \\W&,R, 44Vn \P!V\PR7Vn^\"P$!\"P&!VP44,pVP 'd V^,pVVn\P*!VP\P,R7pR^\P.!^\P0,V,VP, 4, ,pVP3\P,4Vn\7VP(^,^,VVVVP8RVR7VnVe.\P!V4P=^^V4VnMRVnVe/\P!V4P=^^V4Vn R#RVn R#) rTrUrVrWg@@rg?N)r r r rrrr)!super__init__rZr[r\r]r^r_r`rround frame_length hop_lengthrr>float64rar)ceillog2rr'r(cospiastypewindowr<rU mel_filtersreshaperbrc)selfrTrUrVrWrXrYrZr[r\r]r^r_r`rarbrckwargsr hann_arangert __class__s&&&&&&&&&&&&&&&&&, rrj%Gemma3nAudioFeatureExtractor.__init__s(  % ' ( #8    +*&&<#* "4m&E&N OPeM$AF$JKL)2::>$))DIId.?.?$@AA     !OJ$ii 1 1D BFF1ruu9{#:T=N=N#NOOPmmBJJ/ +OOq(1,**!   # " 6 > >q!\ RD  $D   %"$((>":"B"B1a"VD "&D rc<V^8dQhRS[PRS[PRS[S[PS[P3,/#)r waveformattention_maskr)rrtuple)rres"rrrfsH+A+ARZZ+A+AX]^`^h^hjljtjt^tXu+Arc VP^8Xd\P!V^R7pVPR8dYWP\PP !VP !PVP4,,pVPR8wdWP,pVP^,p\VRW0PR7pVPR8dVP'dlVR ,RVP, ,pVR^R13,VPVRRR 13,,, p\P!WV.RR7pM:VR ,VPVRRR13,,, pM VRRR13,pWpP ,p\P"P%WpP&RR7p\P(!V4p \P*!WP,4p \P.!\P0!WP244p VP4eWP4, p VP6eWP6, p V P9^4p VRRVP1,P\:4p WRV P ^,3#) )axisr )r?r@rA.N)nrr$).:Nr"Nr%).r!)rErr,r_randomrandnrCrsrr`rlrNrmr\r] concatenatertfftrfftrabsmatmulrulogr.rarbrcsqueezerd)rwr}r~frame_size_for_unfoldframes_to_processfirst_in_frame rest_in_frameframesstftmagnitude_specmel_spec log_mel_specmel_spectrogrammasks&&& r_extract_spectrogram1Gemma3nAudioFeatureExtractor._extract_spectrograms? ==A ~~hQ7H ;; ++ 0P0W0WX`XfXf0g"ggH  " "c )"9"99H $ 1 1A 5$HAV]l]lm   c !***!27!;sTEUEU?U!V 1#qt) 99^-=-=>vvbjj>>BC    ('*;*;;L    *'*=*==L&..q100188>%?'<'tt:jYZm^`^h^hjr]sCt%<<)< $J3=>:Rbjj"&((: > *J7 8!!1"7 " ! = =~?\?\]LFD44VXXtDLFD  " "6==#< = ' ' -^  0EG[ \&  %?s*F)r_rr^rlrmr`r[rurarZrbrcr\r]rt)i>rTg@@g$@g@_@g@g ףp= ?TTrr gh㈵>NN)longestiSTrNT) __name__ __module__ __qualname____firstlineno____doc__model_input_namesrjrr__static_attributes____classdictcell__ __classcell__)rzres@@rrPrPlsB)V*+@AB'B'H+A+AZ? ? ? rrP)N)r)collections.abcrnumpyr!feature_extraction_sequence_utilsrfeature_extraction_utilsrutilsrrr get_loggerrloggerr<rNrP__all__rhrrrsT $I499   H %:z^&^ #;^ B * *r