+
j63 ^ RI t^RIHt ^RIHt ^RIHtHtH t H
t
^RI H t ] ! 4 ' d ^ RI
t
]! 4 ' d ^ RItRtR
t]
P" ! ]4 t] ! R R7 ! R R ]4 4 tR .tR# ) N)SequenceFeatureExtractor) BatchFeature)
TensorTypeis_librosa_availableis_torch_availablelogging)requiresgh㈵>)backendsc b a a ] tR t^%t oRtRR.tR V 3R lltR
R ltR V3R lR lltRt Vt
V ;t # ) ParakeetFeatureExtractoraB
Constructs a Parakeet 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.
n_fft (`int`, *optional*, defaults to 512):
Size of the Fourier transform.
win_length (`int`, *optional*, defaults to 400):
The window length for the STFT computation.
preemphasis (`float`, *optional*, defaults to 0.97):
A preemphasis filter coefficient. 0.0 means no preemphasis filter.
padding_value (`float`, *optional*, defaults to 0.0):
Padding value used to pad the audio. Should correspond to silences.
input_featuresattention_maskc * < \ S
V ` ! RR VRVRV/VB W0n W@n WPn W`n \ P P W$VRV^, RR7 p \ P ! V 4 P \ P 4 V n
R# ) feature_size
sampling_rate
padding_value slaney)srn_fftn_melsfminfmaxnormN )super__init__
hop_lengthr
win_length preemphasislibrosafiltersmeltorch
from_numpytofloat32 mel_filters) selfr r r r r r r kwargsr( __class__s &&&&&&&&, y/home/luca/screen-checking/.venv/lib/python3.14/site-packages/transformers/models/parakeet/feature_extraction_parakeet.pyr !ParakeetFeatureExtractor.__init__C s wlw-w_lwpvw$
$& oo)) ,S}_`O`go *
!++K8;;EMMJ c
& \ P ! V P R VR7 p\ P ! VV P V P
V P VRRR7 p\ P ! V4 p\ P ! VP ^4 P R4 4 pVP ^4 pV P P V4 pWe, p\ P ! V\ , 4 pVP ^ ^^4 pV# )F)periodicdeviceTconstant)r r windowreturn_complexpad_mode)r$ hann_windowr stftr r view_as_realsqrtpowsumr( r& logLOG_ZERO_GUARD_VALUEpermute)r) waveformr1 r3 r8
magnitudesr( mel_specs &&& r, _torch_extract_fbank_features6ParakeetFeatureExtractor._torch_extract_fbank_featurese s ""4??U6Rzz JJ
''-
ZZ
q 1 5 5b 9:
^^A&
&&))&1 +99X(< <= ##Aq!,r. c < V ^8 d QhRS[ P S[S[, , S[S[ P , , S[S[S[, , , RS[RS[R, RS[S[, R, RS[R, RS[R, RS[R, R S[R, R
S[R, R S[R, R S[R, R
S[/ # )
raw_speech
truncationpad_to_multiple_ofNreturn_tensorsreturn_attention_maskpadding
max_lengthr do_normalizer1 return_token_timestampsreturn) npndarraylistfloatboolintstrr r )format
__classdict__s "r, __annotate__%ParakeetFeatureExtractor.__annotate__ s [
[
JJe,tBJJ/??$tE{BSS[
[
$J [
j(4/ [
$d{
[
t[
$J[
Tz[
Tk[
d
[
"&[
[
r. c
VeR WP 8w dA \ RV P P RV P RV P RV R2 4 hM-\ P RV P P R24 \
V\ P 4 ' d \ P ! V4 pMh\
V\ \ 34 ' dM \
V^ , \ P 4 ' d& V U
u. uF p
\ P ! V
4 NK pp
\
V\ P 4 ;' d \ VP 4 ^8 pV' dY \ VP 4 ^8 d? \ P R V P P R
24 VP! R4 p\
V\ \ 34 pV' dd V F] p
\ V
P 4 ^8 g K \ P R V P P R
24 V
P! R4 p
K_ V' g V' d6 V U
u. uF( qR, P# \ P$ 4 NK* pp
M'VR, P# \ P$ 4 .pV U
u. uF p
\ V
4 NK pp
\' R VR
V/4 pV P) VVVVVRR7 pVP* P- R4 pV P. e \ P0 ! VP ^, VP2 R7 P5 ^ 4 VP6 P5 ^4 8 p\ P8 ! VR, VR, V P. VR RR13, , ,
.^R7 pVP; V( R4 pV P= VV
4 p\ P> ! VP6 V P@ ^, ^, , V P@ ,
V PB 4 p\ P0 ! VP ^, V
R7 R, VR, 8 pVP5 R4 pVV, pVPE ^R7 VP5 R4 , pVP5 ^4 pVV,
^, V, PE ^R7 V^,
P5 R4 , p\ PF ! V4 P5 ^4 pVV,
V\H , , pVV, p\' R VRV/VR7 # u up
i u up
i u up
i )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 Parakeet 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`):
Deprecated. Use `return_attention_mask` instead from which the number of frames can be inferred.
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.z2Only mono-channel audio is supported for input to z;. We will take the mean of the channels to convert to mono.NNNr
audio_lengthspt)rL rM rH rI rJ )r1 )dimr r )data tensor_typer6 )r^ N)r^ :N N)r^ :rd NN)Nr^ )%r
ValueErrorr+ __name__loggerwarning
isinstancerQ rR r$ tensorrS tupleTensorlenshapemeanr& r' r padr
squeezer aranger1 unsqueezer_ cat masked_fillrC floor_divider r r< r: EPSILON) r) rG rH rI rJ rK rL rM r rN r1 rO r* speechis_batched_torchis_batched_sequencer_ batched_speech
padded_inputsr
timemaskfeatures_lengthsr maskinput_features_maskedro variancestds &&&&&&&&&&&&, r, __call__!ParakeetFeatureExtractor.__call__ s H $ 2 22 I$..JaJaIb c))-););(< =))-););(:V:
>%'7_Vc&de !! 1
!
'55==bA ' ||N$8$8$;NDYDYZdd ++55a89H #YY &u(=@P@PSabcehfhehbhSi@i(ijpq N , 7 7 3 GN ;;NFS -- ' '$**/A*= =
JDOO
n&:&:1&=fMgVYijqYrr ''+ . 5 $((Q(/2B2L2LR2PP~~a *T1a7$>CCCJN^abNbMmMmnpMqqjj",,Q/(4/C'MB$ . . '
A I* V ?s U&.U+-U0)r r( r r r )P i> i i g
ףp=
?r )cpu)
FNNNlongestNNNr N) rf
__module__ __qualname____firstlineno____doc__model_input_namesr rC r __static_attributes____classdictcell__
__classcell__)r+ rY s @@r, r r % s6 4 *+;< KD8[
[
[
r. r g p>)r$ r! )numpyrQ !feature_extraction_sequence_utilsr feature_extraction_utilsr utilsr r r r utils.import_utilsr r$ r! rw r>
get_loggerrf rg r __all__r r. r, r s I 4 R R *
H %
'(v
7 v
)v
r &
&r.