+ jX6$^RIt^RIHt^RIHtHt^RIHtH t H t H t ^RI H t Ht^RIHtHtHt^RIHt^RIHt]!4'd ^R IHtHt]P4!]4t!R R ] R R 7t]]!RR7!RR] 444tR.tR#)N) AudioInput) ImageInputmake_nested_list_of_images)MultiModalDataProcessingKwargsProcessorMixinUnpack)PreTokenizedInput TextInput)auto_docstringis_vision_availablelogging)requires) VideoInput)Gemma4ImageProcessorKwargs get_aspect_ratio_preserving_sizec Pa]tRt^"toRRRRR/RRR/R/RR R//tV3R ltR tVtR #) Gemma4ProcessorKwargs text_kwargspaddingTreturn_mm_token_type_ids images_kwargsdo_convert_rgb audio_kwargs videos_kwargsreturn_metadatac&<V^8dQh/S[;R&#)r)r)format __classdict__s"m/home/luca/screen-checking/.venv/lib/python3.14/site-packages/transformers/models/gemma4/processing_gemma4.py __annotate__"Gemma4ProcessorKwargs.__annotate__"s--N)__name__ __module__ __qualname____firstlineno__ _defaults__annotate_func____static_attributes____classdictcell__)r s@r!rr"sH  t &   d  +T2 Ir$rF)total)backendscaa]tRt^1to]tRV3RlV3RllltRV3RlV3RllltRV3RlV3RllltV3RlRlt V3R lR lt V3R lR lt RR lt V3RlRlt ]V3Rl4t]V3RlRl4tRtVtV;t#)Gemma4Processorc,<V^8dQhRS[RS[RS[/#)rimage_seq_lengthaudio_seq_lengthaudio_ms_per_tokenint)rr s"r!r"Gemma4Processor.__annotate__6s+6 6 6 6  6 r$c <W`nVPVnVPVnVPVnVPVnVP RR./4RVnVPVP 4VnWpn Wn \VRR4Vn \VRR4Vn \VRR4Vn\VRR4Vn\ S V`D!R RVR VR VR VR V/V BR#)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_tokenfeature_extractorimage_processor tokenizervideo_processor chat_templater%)r3image_token_id boi_token eoi_token image_tokenadd_special_tokens video_tokenconvert_tokens_to_idsvideo_token_idr4r5getattrr;r<r=r>super__init__) selfr?r@rArBrCr3r4r5kwargs __class__s &&&&&&&&&,r!rNGemma4Processor.__init__6s.!1'66",,",,$00 $$&AK=%QR&'==d>N>NO !1#5%i1A4H"9mTB K> K>  / +   ,  (    r$c<V^8dQhRS[R,RS[S[,S[S[,,S[S[,,RS[RS[/#)rimagesNtextvideosaudio)rr r listrr)rr s"r!r"r8nsR++T!+++d9o=EV@WW+ +  +r$c Z<\SV`!RRVRVRVRV/VBwrr4Ve \V4pV'dDV'g<VUu.uF/pRPVP.\ V4,4NK1 ppV'd&V'gVP .\ V4,pWW43#uupi)rTrUrVrW r%)rMprepare_inputs_layoutrjoinrGlenr<)rOrTrUrVrWrP image_listrQs&&&&&, r!r[%Gemma4Processor.prepare_inputs_layoutns',g&C' ' $' -3' ;@' DJ' #f  /7F $U[\U[zCHHd../#j/ABU[D\ $$%E 2DV** ]s5B(c <V^8dQhRS[S[S[,,R,RS[S[,S[S[,,S[S[,,RS[RS[RS[S[,/#)rrTNrUrVrWrP)rrXr r rrr r)rr s"r!r"r8sl!!T*--4!++d9o=EV@WW! !  ! )* !r$c <\S V`!RRVRV/VBVfVf \R4hVeVPeVPeVP f \R4hVeVUu.uFqfP VP4NK ppVe\V4\V48wd%\R\V4 R\V4 R24hVUu.uFp\V4NK p pWy8wd-\RVP R V R VP R V R 2 4hR#Vf9\V4'd&\R \V4 R VP R24hR#R#R#uupiuupi)rTrUNz+You must provide either `text` or `images`.zUAudio inputs were provided, but the tokenizer does not have an `audio_token` defined.z1Received inconsistently sized batches of images (z ) and text (z).zThe total number of zP tokens in the prompts should be the same as the number of images passed. Found rZz tokens and z images per sample.zFound z. tokens in the text but no images were passed.r%) rMvalidate_inputs ValueErrorr<r=r>countrGr]anysum) rOrTrUrVrWrPsamplen_images_in_textsublistn_images_in_imagesrQs &&&&&, r!rbGemma4Processor.validate_inputss CvCDCFC T R!v;#d)+$KCPVK=XdehimendooqrCI%I&wc'l&"%I#9$.t/?/?.@A""2!31T5E5E4FlSeRffy{: C(8$9$9 S!1231T5E5E4FFtu%: R&Js *#EE#c,<V^8dQhRS[RS[RS[/#)r image_inputs image_idxreturndictr7str)rr s"r!r"r8s(WWWWSVWr$c |VR,V,pVP VPV, VP 2#)num_soft_tokens_per_image)rErGrF)rOrmrnrPnum_soft_tokenss&&&, r!replace_image_token#Gemma4Processor.replace_image_tokens;&'BCIN..!$"2"2_"D!EdnnEUVVr$c,<V^8dQhRS[RS[RS[/#)r video_inputs video_idxrorp)rr s"r!r"r8s#!!!!SV!r$c VR,V,pVR,V,pVPf\PR4VPf^M VPVnVPUu.uF,p\ V^<,4R R\ V^<,4R 2NK. ppRP VUu.uF2q RVP VPV, VP 2NK4 up4p V #uupiuupi)num_soft_tokens_per_videovideo_metadataaGemma4 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:rZ) fpslogger warning_once timestampsr7r\rErIrF) rOryrzrPrumetadataseconds timestamp_strtvideo_replacements &&&, r!replace_video_token#Gemma4Processor.replace_video_tokens&'BCIN 01)< <<    e  &\\1rx|| ]e\o\op\oQXC2 .s31S25Fs4KL\o pHHbo pbo]^s!DNN#D$4$4$F#GGW Xbo p ! q ps 32C768C<c,<V^8dQhRS[RS[RS[/#)r audio_inputs audio_idxrorp)rr s"r!r"r8s( W W W WSV Wr$c NVR,V,p\V4p\^4F8pV^,^, ^,^,pVR,RVp\V4pK: VP VP\ VP 44, VP 2#)input_features_mask:NNrN)r]ranger=r<r7rfr>)rOrrrPmaskr_t_outs&&&, r!replace_audio_token#Gemma4Processor.replace_audio_tokens129= IqAUQY1$q(E9Ve$DD A ..!$"2"2S_"D!EdnnEUVVr$c \PPR/4pVPV4VPRR4;'gVPP pVPRR4;'gVPP pVPRR4;'gVPPpWv^,,p/p Ve.p VFXp \V ^,V ^,VVVR7wrW,pW,pV PW,V^,,4KZ ^.\V4,pV PRV RV/4Ve`\VPR R 4pVUu.uF)pVP\P!V4V4NK+ ppV PR V/4\!R /V B#uupi) a Computes the number of placeholder tokens needed for multimodal inputs with the given sizes. Args: image_sizes (`list[list[int]]`, *optional*): The input sizes formatted as (height, width) per each image. audio_lengths (`list[int]`, *optional*): The lengths of audio inputs in number of samples. Used to dynamically compute per-audio token counts. Returns: `MultiModalData`: A `MultiModalData` object holding number of tokens per each of the provided input modalities, along with other useful data. r patch_sizeNpooling_kernel_sizemax_soft_tokens)heightwidthr max_patchesrnum_image_tokensnum_image_patches sampling_ratei>num_audio_tokensr%)rr*getupdater@rrrrappendr]rLr?_compute_audio_num_tokensnpzerosr)rO image_sizes audio_lengthsrPrrrrr vision_datar image_sizetarget_htarget_w patch_height patch_widthrrlengthrs&&&, r!_get_num_multimodal_tokens*Gemma4Processor._get_num_multimodal_tokenss .77;;ORP V$"&&|T:]]d>R>R>]>]   3T : f fd>R>R>f>f (++,=tDllH\H\HlHl%Q(>>   "! ) %E%a=$Q-) +(; &" (5 &4  '' (BFY[\F\(\]*"#c+&6 6     24DFY[lm n  $$D$:$:OVTM^k ^kTZ..rxx/?O^k      24DE F, ,,  s>/G c&<V^8dQhRS[RS[/#)rrror6)rr s"r!r"r8s&-&-s&-s&-r$c \V4pVPP^,pVPP^,pW5,V, VPP,^,pV^8:d^#^p^^^rpTp \ V4F(p V ^V ,,V, V ,^,p K* \ WP 4#)aPNumber of audio soft tokens, replicating the encoder's seq-length arithmetic. Mirrors Gemma4AudioFeatureExtractor mel framing + the two stride-2 Conv2d subsampling layers in Gemma4AudioSubSampleConvProjection, capped at ``audio_seq_length``. Must match ``audio_mask.sum()`` from the audio tower or vLLM's ``_merge_multimodal_embeddings`` will raise on a length mismatch. Args: audio_waveform: A 1-D array or list containing the raw audio samples. sampling_rate: The sampling rate of the audio waveform in Hz. Returns: The number of audio soft tokens to insert as placeholders. )r]r? frame_length hop_lengthrminr4) rOaudio_waveformr num_samplesframe_size_for_unfoldpad_leftnum_mel_framessscp_num_layers sscp_kernel sscp_stride sscp_paddingrrs &&& r!r)Gemma4Processor._compute_audio_num_tokenss.) !% 6 6 C Ca G))66!;%03HHTMcMcMnMnnqrr Q 12Aq, 'AQ%% 3 CaGA( 1++,,r$c(<\SV`R.,#)mm_token_type_ids)rMmodel_input_names)rOrQs&r!r!Gemma4Processor.model_input_names,sw(,?+@@@r$c0<V^8dQhRS[S[,/#)rro)rXrr)rr s"r!r"r81sJJDIJr$c RR.#)rtr|r%)rOs&r!unused_input_names"Gemma4Processor.unused_input_names0s+-HIIr$) r5r4r<r;r=rEr>rFr3rGrDrIrK)Nii()NNNN)NN)r&r'r(r)rvalid_processor_kwargsrNr[rbrvrrrrpropertyrrr,r- __classcell__)rQr s@@r!r1r11s36 6 p++0!!FWW!!& W W5-n&-&-PAAJJJr$r1)vision) numpyr audio_utilsr image_utilsrrprocessing_utilsrrrr tokenization_utils_baser r utilsr r rutils.import_utilsr video_utilsrimage_processing_gemma4rr get_loggerr&rrr1__all__r%r$r!rs %AXXCAA*%e   H % ,E  ;JnJ JD  r$