fThW>SSKrSSKJr SSKJrJrJrJrJr SSK r SSK J r J r \ "5(aSSKr\ "5(aSSKrSSKJrJr SSKJr SSKJrJrJrJr SS KJrJr "S S \S S 9r"SS\S S 9r"SS\5r S/r!g)N)Path)AnyDictListOptionalUnion)is_soundfile_availableis_torch_available) AudioInputmake_list_of_audio) BatchFeature) AudioKwargsProcessingKwargsProcessorMixinUnpack)PreTokenizedInput TextInputc0\rSrSr%\\\\4\S'Sr g)CsmAudioKwargs(encoded_length_kwargsN) __name__ __module__ __qualname____firstlineno__rrstrr__annotations____static_attributes__r^/var/www/auris/envauris/lib/python3.13/site-packages/transformers/models/csm/processing_csm.pyrr(s#DcN33r!rF)totalcT\rSrSr%\\S'SSSS./SQ/SQ/S QSS .S S .S S0S.rSrg)CsmProcessorKwargs, audio_kwargsTleftF)padding padding_sideadd_special_tokens)r r r- r r- r r-r )r-r-r-r2r-r-r-r-r-r-r.r-)r-r-r-r-r-r-r-r-r-r-r-r-r-r-r-) kernel_sizesstrides dilationsuse_causal_convi])r sampling_ratereturn_tensorspt) text_kwargsr' common_kwargsrN)rrrrrr _defaultsr rr!r"r%r%,sG  ""' !QHJ#' & # +D1Ir!r%c ^\rSrSrSrSS/rS/rSrSrSU4Sjjr \ SS j5r S \ S \ \\\\ \\44S \\4S jrSS\\ \\\\\\4S \\ S\\S\\S \\4 SjjrSrU=r$) CsmProcessorAa Constructs a Csm processor which wraps [`EncodecFeatureExtractor`] and [`PretrainedTokenizerFast`] into a single processor that inherits both the audio feature extraction and tokenizer functionalities. See the [`~CsmProcessor.__call__`] for more information. The preferred way of passing kwargs is as a dictionary per modality, see usage example below. ```python from transformers import CsmProcessor from datasets import load_dataset ds = load_dataset("hf-internal-testing/dailytalk-dummy", split="train") audio = ds[0]["audio"]["array"] processor = CsmProcessor.from_pretrained("eustlb/csm-1b") processor( text=["<|begin_of_text|>[0]What are you working on?<|end_of_text|><|AUDIO|><|audio_eos|><|begin_of_text|>[1]I'm figuring out my budget.<|end_of_text|>"], audio=audio, text_kwargs = {"padding": False}, audio_kwargs = {"sampling_rate": 16000}, common_kwargs = {"return_tensors": "pt"}, ) # this should error out because EncodecFeatureExtractor expects a 24kHz audio :) ``` Args: feature_extractor ([`EncodecFeatureExtractor`]): The feature extractor is a required input. tokenizer ([`PreTrainedTokenizer`, `PreTrainedTokenizerFast`]): The tokenizer is a required input. chat_template (`str`, *optional*): A Jinja template which will be used to convert lists of messages in a chat into a tokenizable string. feature_extractor tokenizer chat_templateEncodecFeatureExtractorPreTrainedTokenizerFastc>[US5(d(SUlURUR5UlO"URUlURUl[US5(d(SUlURUR5UlO"URUlUR Ul[ TU]XUS9 g)N audio_tokenz <|AUDIO|>audio_eos_tokenz <|audio_eos|>)rE)hasattrrIconvert_tokens_to_idsaudio_token_idrJaudio_eos_token_idsuper__init__)selfrCrDrE __class__s r"rPCsmProcessor.__init__js y-00*D "+"A"A$BRBR"SD (44D "+":":D y"344#2D &/&E&EdFZFZ&[D ##,#<#*CsmProcessor.save_audio..s#@qep`aAPSUY{A[A[eps#%zAInvalid input path. Please provide a string, or a list of stringsz5The number of audio and saving paths must be the samer'r:)r ImportErrorr rlrrlisttupleall ValueErrorlen _merge_kwargsr%rUtorchTensorcpufloatnumpysfwrite) rQrfrgrh output_kwargsr'r: audio_valuerns r" save_audioCsmProcessor.save_audios &''OP P#5) kC; / /&-K[4-88S@qep@q=q=q`a a u:[) )TU U**   %^4 $_5 !%5NK+u||44)oo/557==? HHQ] 36r!text output_labelsdepth_decoder_labels_ratioc UR"[4SURR0UD6nUSnUSnUSn U R SS5n U S:wa"[ UR RS35e[U[5(aU/nO=[U[[45(a[S U55(d [ S 5eUV s/sHoRUR5PM n n S n Ub[U5n[!U5n [#U 5S :a/U [#U 5:wa Uc [ S 5e[ S U SU S35eUGbUR S05nUVs/sH"oR$"UR&S40UD6PM$ nnUR)5n/nUHn/nURU;a`UR S 5nURU-nUR+U5 UR-URSS5nURU;aM`SU;a*UR-SUR S 5S5nSU;aM*UR+U5 M UnUR"U40UD6n0nUR/U5 UGbUR SS5 //nnS nU GH*n U S :XaMUR+[0R2"S 55 UR+[4R6"S/55 MWUR+[0R8"UUUU -Vs/sHBn[U[4R:5(aUR=5R?5OUPMD snSS95 UR+[4R6"UUUU -Vs/sHnUR&SPM sn5RASS95 UU - nGM- URB"U40UD6nUR SS5 UR/U5 [ESU55nUVs/sH>n[4RFRHRKUS UUR&S- 4SS9PM@ nn[4RL"US S9US'U(aUSURN:HRQ5n U R&S n!US::a.[4RR"U!5S[UU!SU- -5n"U U"n#OU n#[4RV"USURN:HUSS5n$SU$U#SS2S 4U#SS2S44'U$US'[YUU S 9$s sn fs snfs snfs snfs snf)!a Main method to prepare text(s) and audio to be fed as input to the model. This method forwards the `text` arguments to PreTrainedTokenizerFast's [`~PreTrainedTokenizerFast.__call__`] to encode the text. To prepare the audio, this method forwards the `audio` arguments to EncodecFeatureExtractor's [`~EncodecFeatureExtractor.__call__`]. Please refer to the docstring of the above two methods for more information. Args: audio (`np.ndarray`, `torch.Tensor`, `List[np.ndarray]`, `List[torch.Tensor]`): The audio or batch of audio to be prepared. Each audio can be a NumPy array or PyTorch tensor. text (`str`, `List[str]`, `List[List[str]]`): The sequence or batch of sequences to be encoded. Each sequence can be a string or a list of strings (pretokenized string). If the sequences are provided as list of strings (pretokenized), you must set `is_split_into_words=True` (to lift the ambiguity with a batch of sequences). output_labels (bool, *optional*, default=False): Whether to return labels for training. Indices will be in `[config.audio_token_id, -100, -101]`. - `config.audio_token_id` indicates an audio frame (considering sequence length elements as frames) - `-100` will be ignored in the loss computation - `-101` indicates the audio frame will be used only for the backbone model (using the first codebook token as labels) depth_decoder_labels_ratio (float, *optional*, default=1.0): The ratio of audio frames to keep for the depth decoder labels. return_tensors (`str` or [`~utils.TensorType`], *optional*): If set, will return tensors of a particular framework. Acceptable values are: - `'tf'`: Return TensorFlow `tf.constant` objects. - `'pt'`: Return PyTorch `torch.Tensor` objects. - `'np'`: Return NumPy `np.ndarray` objects. - `'jax'`: Return JAX `jnp.ndarray` objects. Returns: [`BatchFeature`]: A [`BatchFeature`] with the following fields: - **input_ids** -- List of token ids to be fed to a model. Returned when `text` is not `None`. - **input_values** -- List of audio values to be fed to a model. Returned when `audio` is not `None`. - **attention_mask** -- List of indices specifying which tokens should be attended to by the model (when `return_attention_mask=True` or if *"attention_mask"* is in `self.model_input_names` and if `text` is not `None`). - **labels** -- List of labels for the audio frames. Returned when `output_labels=True`. tokenizer_init_kwargsr=r'r>r;Nr<z% only supports `return_tensors='pt'`.c3B# UHn[U[5v M g7frk)rlr)rmts r"ro(CsmProcessor.__call__..s9[VZQR*Q:L:LVZszAInvalid input text. Please provide a string, or a list of stringsrz@No audio were provided, but there are audio tokens in the promptz)The number of audio tokens in each text (z7) should be the same as the number of provided audios (z).rz r-return_attention_mask)axis)dim padding_maskc3># UHoRSv M g7f)rN)shape)rmcut_idxss r"rorSsR=Q..,=Qs)valueinput_values_cutoffs input_ids?iilabels)data tensor_type)-rwr%rD init_kwargspoprurRrrlrrrrsrtcountrIr rvsumrdrcopyappendreplaceupdatenpzerosrxtensor concatenateryrzr|cumsumrCmaxnn functionalpadstackrMnonzerorandpermintwherer)%rQrrfrrrhrr=r'r>r;rn_audio_in_textn_audior audio_arraynum_audio_tokens_listnum_audio_tokens_list_copy expanded_textsample replace_strnum_audio_tokensexpanded_audio_tokenencodingrconcatenated_audioroffsetel audio_inputsmax_lenraudio_frame_idxsn_audio_frames rand_idxsskip_frames_idxsrs% r"__call__CsmProcessor.__call__s~^**  "&.."<"<  $M2 $^4 %o6 &**+;TB T ! 7 788]^_ _ dC 6DTD%=11c9[VZ9[6[6[`a a>BCd774#3#34dC  &u-E%jG  ! #33G(G} !cdd ??PQ229">  $0$4$45Lb$Q !lq%lq]h(():):2)>XBWXlq "%*?)C)C)E &M &&&0'A'E'Ea'H$+/+;+;>N+N(&&';<#^^D,<,>$6+6 H     4d ;792 4 F*a<&--bhhqk:(// bT0BC&--+0'9I*J*JB5?r5<<4P4P 0VX X*J"$ )// U6FU\L\=]%^=]rbhhrl=]%^_ffkmfng%F#+& 112DU UL   ^T 2 KK %R=QRRG!5$ 4H##''1gr@R6R2S[]'^ 4 !$,1;;7KQR+SD' (  $[ 1T5H5H HQQS -33A6N)S0!NN>:;sSSTWqSqAr=st #3I#> #3 [[k!2d6I6I!I4P[K\^bcFEIF#AqD)+;AqD+AA B#DN>BBD$%L&_$s $T>*)U%A UU $AU)rJrNrIrMrk)NNNN)NFr)rrrr__doc__ attributes valid_kwargsfeature_extractor_classtokenizer_classrP staticmethodrdr rrrrrr%rrrrboolr{rr __classcell__)rRs@r"rArAAs!F&{3J#$L7/O  T,$$L 4 43d5d+;&<<= 4+, 4J'+(-69 `CuY(94 ?DQbLccde`C #`C ~ `C %-UO `C +, `C`Cr!rA)"rVpathlibrtypingrrrrrr|rutilsr r rx soundfiler} audio_utilsr r feature_extraction_utilsrprocessing_utilsrrrrtokenization_utils_baserrrr%rA__all__rr!r"rs|  33?94UU 4[4)*hC>hCV  r!