^jg3ddlZddlmZmZddlmZddlmZm Z m Z ddl m Z ddl mZmZmZerddlZej$eZGdd ed ZeGd d e Zd gZy)N) AudioInput make_list_of_audio_chat_template) BatchFeature)ProcessingKwargsProcessorMixinUnpack) TextInput)auto_docstringis_torch_availableloggingc*eZdZddidddddddd Zy ) GlmAsrProcessorKwargspaddingTi> max_length) sampling_ratereturn_attention_maskrptleft)return_tensors padding_side) text_kwargs audio_kwargs common_kwargsN)__name__ __module__ __qualname__ _defaultsw/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/transformers/models/glmasr/processing_glmasr.pyrr&s2 t #%)# #"  Ir rF)totalc eZdZeZ dfd Ze ddeeezde dzde dzde ede f fd Z dde dzdeeezdzde effd Zdd Zde fd Zd ededefdZedeeffd ZedeefdZ ddeeeze zdeeezdzde ede fdZdddZdZdedefdZxZS)GlmAsrProcessorNc||_|j||_||_||_t ||||y)a audio_token (`Optional[str]`, *optional*, defaults to `"<|pad|>`"): Special token used to represent audio inputs in the chat template. default_transcription_prompt (`str`, *optional*, defaults to `"Please transcribe this audio into text"`): Default prompt to use for transcription tasks when applying transcription requests. max_audio_len (`int`, *optional*, defaults to 655): Maximum length of audio sequences in seconds. Audio longer than this will be truncated. 655 gives approximately 8192 tokens, corresponding to the maximum sequence length of the text model. ) chat_templateN) audio_tokenconvert_tokens_to_idsaudio_token_iddefault_transcription_prompt max_audio_lensuper__init__)selffeature_extractor tokenizerr&r'r*r+ __class__s r!r-zGlmAsrProcessor.__init__;sE$''==kJ,H)* *I]Sr Ftextaudio output_labelskwargsreturnc Jd|vr*|ddk7r"t|jjd|rd|d<t|d ||d|}|rM|j d}|dj }d ||d k7<d |||jjk(<||d <t|d S)a0 output_labels (bool, *optional*, default=False): Whether to return labels for training. Returns: [`BatchFeature`]: A dictionary with tokenized text (`input_ids`, `attention_mask`) and audio features (`input_features`, `input_features_mask`). rrz% only supports `return_tensors='pt'`.Treturn_mm_token_type_idsr3r2mm_token_type_ids input_idsirlabels)data tensor_typer) ValueErrorr1rr,__call__popcloner0 pad_token_idr) r.r2r3r4r5 model_inputsr:r<r1s r!r@zGlmAsrProcessor.__call__Ss" v %&1A*Bd*J 7 788]^_ _ 15F- .w'Ie$I&I  , 0 01D E !+.446F-1F$) *<@F6T^^888 9%+L "4@@r c t|d||d||?|.>?O.P *+  #kk,7L*M*Q*QRT*UWij kQUUW k ,0+G+G +V '(U$ 8C#77 PS7T   % %&6 7 8/// ls!HrrrWc0|d|}|j|zS)NrV)r')r.rrrWrVs r!rfz#GlmAsrProcessor.replace_audio_tokens%'(:;IF"222r c t|dgzS)NrT)r,model_input_names)r.r1s r!rzz!GlmAsrProcessor.model_input_namessw(,A+BBBr cdgS)zNInput names returned always by subprocessors but not used in model's `forward`rVr)r.s r!unused_input_namesz"GlmAsrProcessor.unused_input_namess ###r promptc ptt|}trU|Dcgc]J}t|tj r,|j jjn|L}}t|}|dk(r td||jg|z}nt|tr|g|z}nt|ttfr}t||k7rtdt|d|dg}|D]L}||j|j!t|tr|j|Ctdn tdt!||D cgc](\} } dt| trd | d nd | d d | d gdg*} } } |j"| fdddd|Scc}wcc} } w)a  Prepare inputs for automatic speech recognition without manually writing the default transcription prompt. Args: audio (`str`, `list[str]`, `np.ndarray`, `torch.Tensor`, `list[np.ndarray]`, `list[torch.Tensor]`): Audio to transcribe. Strings are interpreted as local paths or URLs and will be loaded automatically by the chat template loader; NumPy arrays and PyTorch tensors are forwarded directly. prompt (`str` or `list[str]`, *optional*): Custom prompt(s) to include in the user turn. A list must be the same length as the batch. When `None`, each sample uses `"Transcribe the input speech."`. **kwargs: Additional keyword arguments forwarded to [`~GlmAsrProcessor.apply_chat_template`] (for example `text_kwargs`, `audio_kwargs`, ...). Returns: [`BatchFeature`]: Processor outputs ready to be passed to [`GlmAsrForConditionalGeneration.generate`]. rz)`audio` must contain at least one sample.z Received z prompt(s) for z$ audio sample(s); counts must match.z'Each prompt must be a string or `None`.z<`prompt` must be a string, a sequence of strings, or `None`.userr3)typepath)rr3r2)rr2)rolecontentT)tokenizeadd_generation_prompt return_dict)listrr isinstancerbTensordetachcpunumpyrGr?r*strtupler^ TypeErrorzipapply_chat_template) r.r3r}r5 audio_itemsel batch_sizepromptsitem prompt_text audio_item conversationss r!apply_transcription_requestz+GlmAsrProcessor.apply_transcription_requests2/33STY3Z.[  grsac 2u||8T299;??,224Z\\sKs% ?HI I >889JFG  $h+G u .6{j( F }OJ>tDYGYZs?c&|j|i|S)z)BC as previous examples used batch_decode)r)r.rr5s r! batch_decodezGlmAsrProcessor.batch_decodest{{D+F++r c@|j}dD]1}|j|s|t|dj}n|jdr|ddj}t|dk\r%|d|dk(r|ddvr|ddj}|S) zi Remove the assistant prefix and surrounding quotes from a decoded transcription string. )z"The spoken content of the audio isz!The transcription of the audio isz!The content of the input audio isN.rUrKr>"'rJ)strip startswithrGendswith)r.r2strippedprefixs r!rz2GlmAsrProcessor._strip_assistant_prefix_and_quotess ::<  F ""6*#CKM288:    S !}**,H x=A (1+""=(1+Q[B["~++-Hr )Nz<|pad|>z&Please transcribe this audio into texti)NF)NN)rL torch.Tensorr6r)N)rrrrvalid_processor_kwargsr-r r rrboolr rr@rrFrQrwdictrXrrfpropertyrzr|rrrr __classcell__)r1s@r!r$r$7s2 %MT0$(%* A$y/)AD Ad{ A ./ A  AAB$(37 dD  d$y/)D0 d)* d(0J(0T3333C49CC$DI$$*.J T#Y+J d3i$&J ./ J  J X*/ ,ssr r$)rnp audio_utilsrrfeature_extraction_utilsrprocessing_utilsrrr tokenization_utils_baser utilsr r r rb get_loggerrr\rr$__all__rr r!rsv,G4HH0@@   H %,E"xnxxv  r