Zj@(SSKrSSKrSSKJr SSKJr SSKJrJ r SSK J r J r J r Jr SSKJrJr SSKJrJrJr SS KJr SS KJr \"5(aS S KJrJr \R:"\5r"S S\ SS9r \\"SS9"SS\ 555r!S/r"g)N) AudioInput) BatchFeature) ImageInputmake_nested_list_of_images)MultiModalDataProcessingKwargsProcessorMixinUnpack)PreTokenizedInput TextInput)auto_docstringis_vision_availablelogging)requires) VideoInput)Gemma4ImageProcessorKwargs get_aspect_ratio_preserving_sizec<\rSrSr%\\S'SSS.SS00SS0S.rSrg ) Gemma4ProcessorKwargs$ images_kwargsT)paddingreturn_mm_token_type_idsdo_convert_rgbreturn_metadata) text_kwargsr audio_kwargs videos_kwargsN)__name__ __module__ __qualname____firstlineno__r__annotations__ _defaults__static_attributes__r!}/root/GenerationalWealth/GenerationalWealth/venv/lib/python3.13/site-packages/transformers/models/gemma4/processing_gemma4.pyrr$s5--(, d +T2 Ir)rF)total)vision)backendsc^\rSrSrSS\S\S\4U4Sjjjr\SS\S-S\\ -\ \-\ \ -S \ S-S \ S-S \ \S \4 S jj5rS\S \4SjrSSjr\U4Sj5rSrU=r$)Gemma4Processor3Nimage_seq_lengthaudio_seq_lengthaudio_ms_per_tokenc >X`lURUlURUlURUlURUlUR SS/05 SUlURUR 5UlXpl Xl [USS5Ul [USS5Ul [USS5Ul[USS5Ul[ T U]D"S UUUUUS.U D6 g) u image_seq_length (`int`, *optional*, defaults to 280): The number of soft tokens per image used for placeholder expansion. audio_seq_length (`int`, *optional*, defaults to 750): The maximum number of audio soft tokens per audio segment. Serves as an upper-bound cap when dynamic audio token counts are computed. audio_ms_per_token (`int`, *optional*, defaults to 40): Milliseconds of audio per output soft token. Used to dynamically compute the number of audio placeholder tokens as ``ceil(duration_ms / audio_ms_per_token)``. The default of 40 comes from the SSCP convolution's 4× time reduction on 10ms frames. additional_special_tokensz <|video|>audio_token_idN audio_token boa_token eoa_token)feature_extractorimage_processor tokenizervideo_processor chat_templater!)r1image_token_id boi_token eoi_token image_tokenadd_special_tokens video_tokenconvert_tokens_to_idsvideo_token_idr2r3getattrr6r7r8r9super__init__) selfr:r;r<r=r>r1r2r3kwargs __class__s r*rIGemma4Processor.__init__6s.!1'66",,",,$00 $$&AK=%QR&'==d>N>NO !1#5%i1A4H"9mTB K> K>  /++'    r)imagestextaudiovideosrKreturnc  ^ ^!UcUcUcUc [S5eUR"[4SURR0UD6n[ U[ 5(aU/nO8[ U[5(d#[ US[ 5(d [S5e0nUGbWURRU5n[U5nUR"U40USD6nURS5n U(d8UVs/sH+nSRUR/[U5-5PM- nn[U5[U5:wa$[S[U5S [U5S 35eU V s/sH+oR URU -UR"3PM- n n [%U 5m [&R("UR5n UV s/sHn [&R*"U U 4S jU 5PM! nn 0nUGbUR,"S'S U0US D6nURS5nUR/S5(dURS5nOUSn/m![1UU5HunnUR2c[4R7S5 UR2cSO UR2UlUR8Vs/sH$n[;US-5SS[;US-5S3PM& nnT!R=SRUVs/sH/nUSUR UR>U-UR"3PM1 sn55 M [%T!5m![&R("UR>5n UV s/sHn [&R*"U U!4SjU 5PM! nn 0nUGbpUR@bURBb URDc [S5e[ U[FRH5(aURJS:XaU/nU(dUR@/[U5-nUR/S05nURL"U40UD6nURLRNnUVs/sHnURQUU5PM nnUV s/sH+oRBUR@U -URD3PM- n n [%U 5m [&R("UR@5nUV s/sHn [&R*"UU 4SjU 5PM! nn USRSS5nUSRSS5nUR"S'SU0USD6n/nWbUR=S 5 UbUR=S!5 UbUR=S"5 U(aURSUUUS#9 U(aURUUS$5US%'[W0UEUEUEUEUS&9$s snfs sn fs sn fs snfs snfs sn fs snfs sn fs sn f)(Nz?Provide at least one of `text`, `images`, `audio`, or `videos`.tokenizer_init_kwargsrzAInvalid input text. Please provide a string, or a list of stringsrnum_soft_tokens_per_image z1Received inconsistently sized batches of images (z ) and text (z).c>[T5$Nnext_replacements_iters r**Gemma4Processor.__call__..s d3D.Er)rQr num_soft_tokens_per_videorvideo_metadataaGemma 4 requires frame timestamps to construct prompts, but the `fps` of the input video could not be inferred. Probably `video_metadata` was missing from inputs and you passed pre-sampled frames. Defaulting to `fps=24`. Please provide `video_metadata` for more accurate results.<02d:c>[T5$rXrY)r\video_replacementss r*r^r_s d3E.Fr)zUAudio inputs were provided, but the tokenizer does not have an `audio_token` defined.rrc>[T5$rXrYr[s r*r^r_s D9J4Kr)rreturn_tensorsrFrOimagevideorP) modalities input_idsmm_token_type_ids)data tensor_typer!), ValueError _merge_kwargsrr< init_kwargs isinstancestrlist TypeErrorr; fetch_imagesrpopjoinrBlenr@rAiterreescapesubr=getzipfpslogger warning_once timestampsintappendrDr7r8r9npndarrayndimr: sampling_rate_compute_audio_num_tokens_check_special_mm_tokenscreate_mm_token_type_idsr)"rJrNrOrPrQrK output_kwargs image_inputsbatched_imagesnum_soft_tokensn replacementspatternprompt video_inputsnum_video_tokensrametadatan_tokensseconds timestamp_strt audio_inputsrranum_audio_tokens audio_patternrir text_inputsactive_modalitiesr]rgs" @@r*__call__Gemma4Processor.__call__ns F7?N//Y-:XYL*../JKOQ_`Q_v$"2"2!3c&k!ABQ_`>"c$i/ GNH[G\\hilmqirhssuv`oo_nZ[~~.t/?/?!/C.DT^^DTU_nLo $\ 2 ii 0 01G]ab]aSYBFF7$EvN]aDb  //`v`A_`L+//0KL ::/00!-!1!12B!C!-.>!?!# &).:J&K"(<<''', &.\\%9rx|| X`WjWj!WjGs7b=)#.aGbL0A#/FGWj!#))HHgtugtbcA3a/0@0@80K/LT^^L\]gtu'L&"&&8!9 ii 0 01G^bc^bTZBFF7$FO^bDc  '4>>+AT^^E[ k %,,q(()CJ6),,^R@L11%H<HL 22@@MZ_`Z_UV > >q- PZ_ `_op_oZ[~~.t/?/?!/C.DT^^DTU_oLp $\ 2 IId&6&67McghcgY_BFF=*KVTcgDh&}599:JDQ#0#?#C#CD^`e#f nnO$O- 2NO     $ $W -    $ $W -    $ $W -   ) )$ HY ) Z #/3/L/L[YdMe/fK+ ,PKP<P<P<P&  Gap c2! v d4 apis662W*2W &W4+W#;6W(/&W- W22W7%&W<rc0[U5n[[US-S- 55n[[US-S- 55nUS-nUS-nX7-nX- U-S-n U S::agU n [S5Hn U S-n U S- S-S-n M [ XR 5$)aCompute the number of audio soft tokens for a single waveform. Replicates the exact sequence-length arithmetic of the audio encoder so that the processor inserts the correct number of placeholder tokens. The computation mirrors: 1. Mel framing via ``_unfold`` in ``Gemma4AudioFeatureExtractor`` 2. Two ``Conv2d`` subsampling layers in ``Gemma4AudioSubSampleConvProjection`` (each: kernel=3, stride=2, semicausal padding top=1, bottom=1) The result is capped at ``self.audio_seq_length`` (the configured maximum). Args: audio_waveform: A 1-D numpy array or list containing the raw audio samples. sampling_rate: The sampling rate of the audio waveform in Hz. Returns: The number of audio soft tokens to insert as placeholders. g4@g@@g$@rrr)r{rroundrangeminr2) rJaudio_waveformr num_samples frame_length hop_lengthframe_size_for_unfoldpad_leftpadded_samplesnum_mel_framesrr\t_paddeds r*r)Gemma4Processor._compute_audio_num_tokenss(.) 5!5!>?@ }t3f<=> ,q 0  1$$/(@ZORSS Q  qA1uHA!#a'A 1++,,r)c D[RRS05nURU5 URSS5=(d URR nURSS5=(d URR nURSS5=(d URRnXvS--n0n Ubd/n UH9n [U SU SUUUS 9upX-nX-nU RX-US--5 M; S/[U5-nU RU US .5 Ub`[URS S 5nUVs/sH)nUR[R"U5U5PM+ nnU RS U05 [!S0U D6$s snf)a Computes the number of placeholder tokens needed for multimodal inputs with the given sizes. Args: image_sizes (`list[list[int]]`, *optional*): The input sizes formatted as (height, width) per each image. audio_lengths (`list[int]`, *optional*): The lengths of audio inputs in number of samples. Used to dynamically compute per-audio token counts. Returns: `MultiModalData`: A `MultiModalData` object holding number of tokens per each of the provided input modalities, along with other useful data. r patch_sizeNpooling_kernel_sizemax_soft_tokensrrr)heightwidthr max_patchesr)num_image_tokensnum_image_patchesri>rr!)rr'rupdater;rrrrrr{rGr:rrzerosr)rJ image_sizes audio_lengthsrKrrrrr vision_datar image_sizetarget_htarget_w patch_height patch_widthrrlengthrs r*_get_num_multimodal_tokens*Gemma4Processor._get_num_multimodal_tokens&s .77;;ORP V$"&&|T:]d>R>R>]>]   3T : fd>R>R>f>f (++,=tDlH\H\HlHl%Q(>>   "! ) %E%a=$Q-) +(; &" (5 &4  '' (BFY[\F\(\]*"#c+&6 6    4D[lm n  $$D$:$:OVTM^k ^kTZ..rxx/?O^k      24DE F, ,,  s 0Fc>[TU]nUVs/sH nUS;dM UPM nnURb?URRnURUVs/sH o"U;dM UPM sn5 US/-$s snfs snf)N)rUr`rn)rHmodel_input_namesr:extend)rJrnamefeature_extractor_input_namesrLs r*r!Gemma4Processor.model_input_names]s!G5* )UU )    ! ! -,0,B,B,T,T )  $ $7T%v7TtduXud7T%v w $7#888 &ws A8A8 A=%A=) r3r2r7r6r8r@r9rAr1rBr?rDrF)Nii()NNNN)NN)r"r#r$r%rrIrrr r rvrrr rrrrrpropertyrr( __classcell__)rLs@r*r/r/3s # #"$6 6 6  6 6 p%)Z^#'$( F T!F ++d9o=EV@WWF D F T! F ./ F  F F P--s--s--^5-n 9 9r)r/)#r}numpyr audio_utilsrimage_processing_utilsr image_utilsrrprocessing_utilsrr r r tokenization_utils_baser r utilsrrrutils.import_utilsr video_utilsrimage_processing_pil_gemma4rr get_loggerr"rrr/__all__r!r)r*rs %2AXXCAA*%i   H % ,E  ;v9nv9 v9r  r)