cCi[3 S SK JrJr S SKrS SKrSSKJr SSKJ r SSK
J r J r J
r
SSKJr \ " 5 ( a S SKrSrS r\
R& " \5 r\" S
S 9 " S S
\5 5 rS
/rg) )OptionalUnionN )SequenceFeatureExtractor) BatchFeature)
TensorTypeis_librosa_availablelogging)requiresgh㈵>g p>)torchlibrosa)backendsc 0 ^ \ rS rSrSrSS/r SU 4S jjrSS jr SS\\ R \ \ \ \ R \ \ \ 4 S\
S \\ S
\\\\4 S \\
S \\ S
\\ S\\ S\\
S\\ S\\
S\4S jjrSrU =r$ )ParakeetFeatureExtractor% aB
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 > [ T
U ] " SXUS.UD6 X0l X@l XPl X`l [ R R X$USUS- SS9n [ R " U 5 R [ R 5 U l
g )N) feature_size
sampling_rate
padding_value slaney)srn_fftn_melsfminfmaxnorm )super__init__
hop_lengthr
win_length preemphasisr
filtersmelr
from_numpytofloat32 mel_filters) selfr r r$ r r% r& r kwargsr, __class__s r/home/james-whalen/.local/lib/python3.13/site-packages/transformers/models/parakeet/feature_extraction_parakeet.pyr# !ParakeetFeatureExtractor.__init__C s wl_lwpvw$
$& oo)) ,S}_`O`go *
!++K8;;EMMJ c
[ R " U R SUS9n[ R " UU R U R
U R USSS9n[ R " U5 n[ R " UR S5 R S5 5 nUR S5 nU R R U5 nXe- n[ R " U[ - 5 nUR SSS 5 nU$ )
NF)periodicdeviceTconstant)r$ r% windowreturn_complexpad_moder r )r hann_windowr% stftr r$ view_as_realsqrtpowsumr, r* logLOG_ZERO_GUARD_VALUEpermute)r- waveformr5 r7 r=
magnitudesr, mel_specs r0 _torch_extract_fbank_features6ParakeetFeatureExtractor._torch_extract_fbank_featurese s ""4??U6Rzz JJ
''-
ZZ
q 1 5 5b 9:
^^A&
&&))&1 +99X(< <= ##Aq!,r2
raw_speech
truncationpad_to_multiple_ofreturn_tensorsreturn_attention_maskpadding
max_lengthr do_normalizer5 return_token_timestampsreturnc
UbP XR :w a@ [ SU R R SU R SU R SU S3 5 eO-[ R SU R R S35 [
U[ R 5 ( a [ R " U5 nOc[
U[ [ 45 ( aH [
US [ R 5 ( a& U V
s/ s H n
[ R " U
5 PM nn
[
U[ R 5 =( a [ UR 5 S
: nU( aW [ UR 5 S : a> [ R S U R R S
35 UR! S5 n[
U[ [ 45 nU( ab U H\ n
[ U
R 5 S
: d M [ R S U R R S
35 U
R! S5 n
M^ U( d U( a6 U V
s/ s H( oSS2S4 R# [ R$ 5 PM* nn
O'USS2S4 R# [ R$ 5 /nU V
s/ s H n
[ U
5 PM nn
[' UUS.5 nU R) UUUUUSS9nUR* R- S5 nU R. b [ R0 " UR S
UR2 S9R5 S 5 UR6 R5 S
5 : n[ R8 " USS2SS
24 USS2S
S24 U R. USS2SS24 - -
/S
S9nUR; U) S5 nU R= UU
5 n[ R> " UR6 U R@ S - S - - U R@ -
U RB 5 n[ R0 " UR S
U
S9SSS24 USS2S4 : nUR5 S5 nUU- nURE S
S9UR5 S5 - nUR5 S
5 nUU-
S - U- RE S
S9US
-
R5 S5 - n[ RF " U5 R5 S
5 nUU-
U[H - - nUU- n[' UUS.US9$ s sn
f s sn
f s sn
f )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.r r; r z2Only mono-channel audio is supported for input to z;. We will take the mean of the channels to convert to mono.r: )r
audio_lengthspt)rO rP rK rL rM )r5 )dimr )r r )data tensor_type)%r
ValueErrorr/ __name__loggerwarning
isinstancenpndarrayr tensorlisttupleTensorlenshapemeanr* r+ r padr squeezer& aranger5 unsqueezerV cat masked_fillrH floor_divider r$ rA r? EPSILON) r- rJ rK rL rM rN rO rP r rQ r5 rR r. speechis_batched_torchis_batched_sequencerV batched_speech
padded_inputsr timemaskfeatures_lengthsr maskinput_features_maskedrh variancestds r0 __call__!ParakeetFeatureExtractor.__call__ s H $ 2 22 I$..JaJaIb c))-););(< =))-););(:V:
>%Vc&de !! 1
!
'55==bA ' ||N$8$8$;NDYDYZdd ++55a89H #YY 2A2&q!"u(=@P@PSabcehfhehbhSi@i(ijpq N , 7 7 3 GN ;;NFS -- ' '$**/A*= =
JDOO
n&:&:1&=fMdTUgVYijkmqjqYrr ''+ . 5 $((Q(/2B2L2LR2PP~~a *T1a7$>CCCJN^abNbMmMmnpMqqjj",,Q/(4/C'MB$ "0"0 '
A I* V ?s 8 S7/S< T)r$ r, r r& r% )P i> i i g
ףp=
?r )cpu)
FNNNlongestNNNr N)r\
__module__ __qualname____firstlineno____doc__model_input_namesr# rH r r` ra rc floatboolr intstrr r r| __static_attributes__
__classcell__)r/ s @r0 r r % s= 4 *+;< KD> !,0;?04!*$('+'+ %26[
"**d5k4
3CT$u+EVVW[
[
%SM [
!sJ!78 [
(~
[
# [
SM[
}[
tn[
[
"*$ [
[
[
r2 r )typingr r numpyr` r !feature_extraction_sequence_utilsr feature_extraction_utilsr utilsr r r
utils.import_utilsr r
rp rC
get_loggerr\ r] r __all__r! r2 r0 r s| # I 4 > > *
H %
'(v
7 v
)v
r &
&r2