Zj9SSKrSSKrSSKJrJr SSKJr SSKJ r J r J r SSK J r SSKJrJr \"5(aSSKr\R$"\5r"SS \ S S 9r"S S \ 5rS /rg)N) AudioInputmake_list_of_audio) BatchFeature)ProcessingKwargsProcessorMixinUnpack) TextInput)is_torch_availableloggingc2\rSrSrSS0SSSS.SSS .S .rS rg ) GlmAsrProcessorKwargs(paddingTi> max_length) sampling_ratereturn_attention_maskrptleft)return_tensors padding_side) text_kwargs audio_kwargs common_kwargsN)__name__ __module__ __qualname____firstlineno__ _defaults__static_attributes__r}/root/GenerationalWealth/GenerationalWealth/venv/lib/python3.13/site-packages/transformers/models/glmasr/processing_glmasr.pyrr(s2 t #%)# #"  Ir"rF)totalc ^\rSrSrSrSU4SjjrSSjrSrSrSS \ \ \ -S \ S-S \ S-S \ \S\4 Sjjr\S\ \4Sj5rSS \\ \-\ -S\\ \-S-S \ \S\4SjjrS S.SjrSrS \S\4SjrSrU=r$)GlmAsrProcessor9aA Constructs an GlmAsr processor which wraps an GlmAsr feature extractor and an GlmAsr tokenizer into a single processor. [`GlmAsrProcessor`] offers all the functionalities of [`WhisperFeatureExtractor`] and [`Qwen2TokenizerFast`]. See the [`~GlmAsrProcessor.__call__`] for more information. Args: feature_extractor ([`WhisperFeatureExtractor`]): The feature extractor is a required input. tokenizer ([`Qwen2TokenizerFast`]): The tokenizer is a required input. chat_template (`Optional[str]`, *optional*): The Jinja template to use for formatting the conversation. If not provided, the tokenizer's default chat template will be used. 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. Ncr>X@lURU5UlXPlX`l[ TU]XUS9 g)N) chat_template) audio_tokenconvert_tokens_to_idsaudio_token_iddefault_transcription_prompt max_audio_lensuper__init__)selffeature_extractor tokenizerr)r*r-r. __class__s r#r0GlmAsrProcessor.__init__Rs=''==kJ,H)* *]Sr"returncdSnSHup4nUSU--US- - S- U-S-nM X- U-S-nU$)N))rr9)r9rr:r9r)r1 audio_lengths merge_factorr kernel_sizestride num_tokenss r#_get_audio_token_length'GlmAsrProcessor._get_audio_token_lengthasY ,B (G&*Q[8K!OLqPU[[^__M-C$2|CaG r"c[R"[R"URS5U5Vs/sHoDR5PM sn5nUR U5n[ R "[ R"UR55n[U5H(upURURU -X5X'M* U$s snf)N) torchstacksplitsumr@recompileescaper* enumeratesub) r1text padding_maskper_sample_windowssr;audio_tokens_lengthsaudio_token_patterni audio_lengths r#_expand_audio_tokens$GlmAsrProcessor._expand_audio_tokensis ekk,BRBRSUBVXj6k$l6kUUW6k$lm #;;MJ jj43C3C)DE()=>OA)--d.>.>.MtwWDG ? %msCcXR:H$N)r,)r1 input_idss r#_get_audio_tokens_mask&GlmAsrProcessor._get_audio_tokens_maskqs////r"FrMaudio output_labelskwargsc UR"[4SURR0UD6nUSnUSnUR S5nUS:wa"[ UR RS35e[U[5(aU/nO=[U[[45(a[SU55(d [ S5e0n UGb[U5n[U5[U5:wa$[ S [U5S [U5S 35e[US UR R"-5n [UR$UR R"-5n /n /n UHn[UR&S 5n[)SX-S- U -5nUU :a;[*R-SXS - SSUR$SUR$S35 U nU R/U5 [1UUU -5n[3U5H.nUU -n[1US-U -U5nU R/UUU5 M0 M UR "U 40UD6n U R5S5nUU S'UR7UUU 5nUR"U40UD6n0UEU EnU(aHUSR95nSUUR;U5'SUUURR<:H'UUS'[?UUS9$)a Main method to prepare one or several text sequence(s) and audio waveform(s) for the model. This method expands `` placeholders in the text based on the post-pool frame counts of the audio windows, then tokenizes the provided strings as-is, and extracts log-mel features with [`WhisperFeatureExtractor`]. If `audio` is `None`, no audio processing is performed and the text is tokenized as-is (LM-only behavior). Args: text (`str` or `list[str]`): Input sequence or batch of sequences. audio (`np.ndarray` or `list[np.ndarray]`): Input audio or batch of audios as NumPy arrays. If provided, there must be as many `text` inputs as `audio` inputs. 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`). tokenizer_init_kwargsrrrrz% only supports `return_tensors='pt'`.c3B# UHn[U[5v M g7frX isinstancestr).0ts r# +GlmAsrProcessor.__call__..s9[VZQR*Q:L:LVZzAInvalid input text. Please provide a string, or a list of stringszGot z text but z audios; they must match 1:1.rrr9zAudio duration (z.1fz s) exceeds zs; truncating to first zs.attention_maskinput_features_maskrYilabels)data tensor_type) _merge_kwargsrr3 init_kwargsget ValueErrorr4rrcrdlisttupleallrlenintr2 chunk_lengthr.shapemaxloggerwarningappendminrangepoprUclonerZ pad_token_idr)r1rMr\r]r^ call_kwargsrrr audio_inputs window_size max_windowsrO flat_chunksaudio_el n_samplesn_wintime_caprSstartendrN text_inputsrmrls r#__call__GlmAsrProcessor.__call__ts!:(( ! "&.."<"<  "-0 ">2 $)9: T ! 7 788]^_ _ dC 6DTD%=11c9[VZ9[6[6[`a a  &u-E4yCJ& 4D {*SZLHe!fggl?;d>T>T>a>aabKd00D4J4J4W4WWXK,. ,.K!q 12 A 7! ; KL;&NN*9O7T+TUX*YYdeiewewdxyPQUQcQcPddfg(E"))%0y%+*=>uA OEq1u 3X>C&&xc':;&"" 11+NNL'++,<=L2>L. /,,TF4..v6 7<@F6T^^888 9#DN>BBr"cURRnURRn[[R X-S/-55$)Nrk)r3model_input_namesr2rsdictfromkeys)r1 tok_names fea_namess r#r!GlmAsrProcessor.model_input_namessBNN44 **<< DMM)"7;P:Q"QRSSr"promptc ~[U[5(aU/nO[U[[45(a*U(a#[ SU55(a [U5nO[[ U55n[ 5(a]UVs/sHPn[U[R5(a,UR5R5R5OUPMR nn[U5nUS:Xa [S5eUcUR/U-nO[U[5(aU/U-nO[U[[45(a[U5U:wa[S[U5SUS35e/nUHTnUcURUR5 M#[U[5(aURU5 MK[!S5e O [!S5e[#Xt5V V s/sH-upS [U [5(aS U S .OS U S .S U S./S./PM/ n n n UR$"U 4SSSS.UD6$s snfs sn n f)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`]. c3B# UHn[U[5v M g7frXrb)reels r#rg>GlmAsrProcessor.apply_transcription_request..s?d^cXZ 2s@S@S^crirz)`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`.userr\)typepath)rr\rM)rrM)rolecontentT)tokenizeadd_generation_prompt return_dict)rcrdrsrtrurr rDTensordetachcpunumpyrvrrr-r} TypeErrorzipapply_chat_template) r1r\rr^ audio_itemsr batch_sizepromptsitem prompt_text audio_item conversationss r#apply_transcription_request+GlmAsrProcessor.apply_transcription_requests/2 eS ! !38'K e} - -%C?d^c?d889JFG  $ $h+G u . .6{j( F }OJ>tDQXGYZsAc&UR"U0UD6$)z)BC as previous examples used batch_decode)r)r1rr^s r# batch_decodeGlmAsrProcessor.batch_decode7s{{D+F++r"c`UR5nSH7nURU5(dMU[U5SR5n O URS5(aUSSR5n[U5S:a(USUS:XaUSS;aUSSR5nU$) zY 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.rCr:r>"'r9)strip startswithrvendswith)r1rMstrippedprefixs r#r2GlmAsrProcessor._strip_assistant_prefix_and_quotes;s ::< F ""6**#CKM288:    S ! !}**,H x=A (1+""=(1+Q[B["~++-Hr")r*r,r-r.)Nz<|pad|>z&Please transcribe this audio into texti)r; torch.Tensorr6r)NFrX)rrrr__doc__r0r@rUrZr rsrboolr rrrpropertyrdrrrrrr! __classcell__)r4s@r#r&r&9s,8%M T0 $(%* ]C$y/)]CD ]Cd{ ]C ./ ]C  ]C~T49TT*.O T#Y+O d3i$&O ./ O  O b*/ ,ssr"r&)rHrnp audio_utilsrrfeature_extraction_utilsrprocessing_utilsrrr tokenization_utils_baser utilsr r rD get_loggerrr{rr&__all__rr"r#rsg, 94HH00   H %,E"XnXv  r"