+
QV-j> ^ RI t ^ RIHtHt ^ RIHt ^ RIHt ^ RIH t H
t
H t H t H
t
^ RIt^ RIt^RIHt ^RIHt ^RIHt ^R IHtHtHt ^R
I H t H t ] ' d
^ R I H t ^R I!H"t" ] ! R
R4 4 t#R R lt$R# ) N) dataclassfield)BytesIO)Path)
TYPE_CHECKINGAnyClassVarOptionalUnion)config)DownloadConfig)
array_cast)
is_local_path
is_remote_urlxopen)no_op_if_value_is_nullstring_to_dict AudioDecoder) FeatureTypec > a ] tR t^t o RtRtRtRtRt] ! RRR7 t
Rt ] P ! R] P ! 4 R] P ! 4 /4 t] ! R RRR 7 tR
tV 3R lR ltRV 3R
lR lltV 3R lR ltV 3R lR ltRV 3R lR lltV 3R ltRtV tR# )Audioa[ Audio [`Feature`] to extract audio data from an audio file.
Input: The Audio feature accepts as input:
- A `str`: Absolute path to the audio file (i.e. random access is allowed).
- A `pathlib.Path`: path to the audio file (i.e. random access is allowed).
- A `dict` with the keys:
- `path`: String with relative path of the audio file to the archive file.
- `bytes`: Bytes content of the audio file.
This is useful for parquet or webdataset files which embed audio files.
- A `dict` with the keys:
- `array`: Array containing the audio sample
- `sampling_rate`: Integer corresponding to the sampling rate of the audio sample.
- A `torchcodec.decoders.AudioDecoder`: torchcodec audio decoder object.
Output: The Audio features output data as `torchcodec.decoders.AudioDecoder` objects, with additional keys:
- `array`: Array containing the audio sample
- `sampling_rate`: Integer corresponding to the sampling rate of the audio sample.
Args:
sampling_rate (`int`, *optional*):
Target sampling rate. If `None`, the native sampling rate is used.
num_channels (`int`, *optional*):
The desired number of channels of the samples. By default, the number of channels of the source is used.
Audio decoding will return samples with shape (num_channels, num_samples)
Currently `None` (number of channels of the source, default), `1` (mono) or `2` (stereo) channels are supported.
The `num_channels` argument is passed to `torchcodec.decoders.AudioDecoder`.
decode (`bool`, defaults to `True`):
Whether to decode the audio data. If `False`,
returns the underlying dictionary in the format `{"path": audio_path, "bytes": audio_bytes}`.
stream_index (`int`, *optional*):
The streaming index to use from the file. If `None` defaults to the "best" index.
Example:
```py
>>> from datasets import load_dataset, Audio
>>> ds = load_dataset("PolyAI/minds14", name="en-US", split="train")
>>> ds = ds.cast_column("audio", Audio(sampling_rate=44100, num_channels=2))
>>> ds[0]["audio"]
>>> audio = ds[0]["audio"]
>>> audio.get_samples_played_in_range(0, 10)
AudioSamples:
data (shape): torch.Size([2, 110592])
pts_seconds: 0.0
duration_seconds: 2.507755102040816
sample_rate: 44100
```
NTF)defaultreprdictbytespath)r initr c V P # N)pa_type)selfs &h/Volumes/fast/ai/experiments/ui-tars-smoke/.venv/lib/python3.14/site-packages/datasets/features/audio.py__call__Audio.__call__] s || c F < V ^8 d QhRS[ S[S[S[S[R3, RS[/# ) valuer return)r strr bytearrayr )format
__classdict__s "r# __annotate__Audio.__annotate__` s0 B BE#ui~*U$V B[_ Br& c ^ RI p^ RIHp Tf \ R4 h\
P ' d ^ RIHp MRp\ T\ 4 ' d RRRT/# \ T\ 4 ' d RRR\ TP 4 4 /# \ T\ \ 34 ' d RTRR/# Te \ Y4 ' d \ T4 # RT9 d \! 4 pT! TP# TR, P% \& P( 4 4 TR , R
7 P+ TR T P, R 7 RTP/ 4 RR/# TP1 R4 Ee \2 P4 P7 TR, 4 ' EdZ TR, P9 R
4 ' Ed& TP1 R 4 f \; R4 hTP1 R4 ' dS \& P< ! TR, \& P> R7 P% \& P( 4 R, pMD\& P@ ! TR, RRR7 P% \& P( 4 R, p\! 4 pT! TP# T4 TR , R
7 P+ TR T P, R 7 RTP/ 4 RR/# RRRTP1 R4 /# TP1 R4 f TP1 R4 e% RTP1 R4 RTP1 R4 /# \ RT R24 h \ d p\ R4 ThRp?ii ; i)zEncode example into a format for Arrow.
Args:
value (`str`, `bytes`,`bytearray`,`dict`, `AudioDecoder`):
Data passed as input to Audio feature.
Returns:
`dict`
N AudioEncoderO>OlPV__ .= =
YYv
*rww~~eFm/L/LV}%%e,,99_-5"#ghh99W%%"$--gbhh"O"V"VWYWaWa"bej"jK"$))E&M3"O"V"VWYWaWa"bej"jK U - -k :oH^_ll5t7H7H m !2FDAA vuyy/@AA
YYw
+uyy/@/LUYYw /69JK K ghmgnnop
g g \]cf f gs
M M MM c
h < V ^8 d QhRS[ RS[S[ S[S[S[S[R3, 3, , RR/# )r( r) token_per_repo_idNr* r )r r
r+ r bool)r- r. s "r# r/ r0 s@ 9 99.6tCsDRVAW token (`bool` or `str`)
Returns:
`torchcodec.decoders.AudioDecoder`
r zw 9SuV}eGn5Z_`fZgimYn token to fetch the files bytes.
local_files (`bool`, defaults to `True`)
Whether to embed local files data in the array
remote_files (`bool`, defaults to `True`)
Whether to embed remote files data in the array.
E.g. files with paths that start with hf:// or https://
Returns:
`pa.StructArray`: Array in the Audio arrow storage type, that is
`pa.struct({"bytes": pa.binary(), "path": pa.string()})`.
Nc < V P R 4 R, pVP \ P 4 ' d \ P M\ P
p\
W4 pVe SP VR, 4 MRp\ VR7 p\ V RVR7 ;_uu_ 4 pVP 4 uuRRR4 # + ' g i R# ; i)rc Nrd re rg rh rj ) rn ro r rp rq rr r rQ r
r read)r rv rw rx rf ri ry r_ s & r#
path_to_bytes*Audio.embed_storage..path_to_bytes3 s D)"-J+5+@+@ASAS+T+T''Z`ZvZv
!/z C K\Kh%))*;I*FGnrE ,5 9OtT?CCqvvx DCCCs *C C r r r r )r r r5 to_pylistr r r r rR r basenamer r r r r r! )
r" r r_ r r r r r r r s
&&f&& r#
embed_storageAudio.embed_storage s 0 $ "
hh !**,
-A = z)%-& *B*B YfghiogpYqYq "!F), 7 -
XX $MM& 1 ; ; =
>D # $
d(;(; R_`dReRe GG$$T*
>
..,,k-FRXHY`g`o`o`q,r'<<007
s# F> F>56F>8G G"8Gc < V ^8 d Qh/ S[ S[, ;R& S[;R& S[ S[, ;R& S[ S[, ;R& S[ S[, ;R& S[S[, ;R& S[S[, ;R& S[;R& # ) r( r6 rl r; rb idr= r! _type)r
intr` r+ r r )r- r. s "r# r/ r0 s x C='y z
{ | 3-&} ~ 3-& @
7A D C=!E F c]SG H ?I r& r )NTT)__name__
__module__ __qualname____firstlineno____doc__r6 rl r; rb r r r= r structr r r! r r$ r\ rz r r r __annotate_func____static_attributes____classdictcell__)r. s @r# r r s 8t $(MF"&L"&L d7B!E YYfbiik'R SGwU?E B BH9 9v
,1 ,1\C1 C1C r& r c ( V ^8 d QhRRR\ /# )r( ru r r* )r )r- s "r# r/ r/ ^ s : :> :d :r& c \ V R 4 ' d
V P # ^ RIHp T P 4 p\
4 pTP P ^ , pT! TP P 4 TP R7 P TRTR7 RTP 4 RR/# \ d p\ R4 ThRp?ii ; i) rs r2 r4 Nr7 r9 r: r r )
hasattrrs rC r3 rD get_all_samplesr datashapecpur8 rO rP )ru r3 rY samplesrZ r; s & r# rJ rJ ^ s um$$ g 8 '') ||))!, W\\%%'W5H5HIVV 5| W
*FD99 g \]cf f gs B/ / C : CC )%rR dataclassesr r ior pathlibr typingr r r r
r numpyrM pyarrowr r download.download_configr
tabler utils.file_utilsr r r utils.py_utilsr r rG r r r r rJ r r& r# r s\ ( @ @ 5 B B C 0% C1 C1 C1L
:r&