o Zh<@sddlZddlmZmZmZddlZddlmZddl m Z m Z ddl m Z mZmZmZddlmZmZddlmZGd d d e ZGd d d ed dZGdddeZdgZdS)N)ListOptionalUnion) BatchFeature) ImageInputmake_nested_list_of_images) ImagesKwargsProcessingKwargsProcessorMixinUnpack)PreTokenizedInput TextInput) to_py_objc@sJeZdZUeeed<eeed<eeed<eeed<eeed<dS)Gemma3ImagesKwargsdo_pan_and_scanpan_and_scan_min_crop_sizepan_and_scan_max_num_crops"pan_and_scan_min_ratio_to_activateZdo_convert_rgbN)__name__ __module__ __qualname__rbool__annotations__intfloatrr[/var/www/auris/lib/python3.10/site-packages/transformers/models/gemma3/processing_gemma3.pyrs    rc@s.eZdZUeed<ddiddddddZd S) Gemma3ProcessorKwargs images_kwargspaddingFg333333?)rrrr) text_kwargsrN)rrrrr _defaultsrrrrr$s  rF)totalc seZdZddgZddgZdZdZ  ddeffd d Z    dd e d e e e e e e e fd eedefddZddZddZeddZZS)Gemma3Processorimage_processor tokenizer chat_templateimage_seq_lengthZAutoImageProcessorZ AutoTokenizerNr!c sh||_|j|_|j|_|j|_d|jg|}d|j||jd|_tjd|||d|dS)Nz )r'r(r)r) r*image_token_id boi_tokenZ image_tokenjoinZ eoi_tokenfull_image_sequencesuper__init__)selfr'r(r)r*kwargsZimage_tokens_expanded __class__rrr19s zGemma3Processor.__init__imagestextr3returnc sp|dur |dur tdjtfdjji|}t|tr"|g}nt|ts2t|dts2tdi}|durt|}j |fi|d}|sRfdd|D}t |t |kritdt |d t |d t | d fd d|D} t t||| D]a\} \} }d dtj| D} t |t | krtdt | dt |dttt| D]-\} }| rdjddjg| }| d||| |t jd} | || <qqfdd|D}|d dd}jd!d|i|dddi}j||dgd|d}t|d}d||jk<dd|D}||d<ti|||d S)"Nz+Provide at least one of `text` or `images`.Ztokenizer_init_kwargsrzAInvalid input text. Please provide a string, or a list of stringsrcs"g|] }djgt|qS) )r.r-len.0r6r2rr ls"z,Gemma3Processor.__call__..z1Received inconsistently sized batches of images (z ) and text (z). num_cropscs&g|]}fddtt|DqS)csg|]}dqS)r)pop)r<_r?rrr>usz7Gemma3Processor.__call__...)ranger:r;rBrrr>us&cSsg|]}|qSr)start)r<mrrrr>wszPrompt contained z image tokens but received z images.zHere is the original image z0 and here are some crops to help you see better r9csg|] }|jjqSr)replacer-r/)r<promptr=rrr>sr#return_tensorsr7npimage)Z modalitiesZ input_idscSsi|] \}}||qSr)tolist)r<kvrrr sz,Gemma3Processor.__call__..token_type_ids)dataZ tensor_typer) ValueErrorZ _merge_kwargsrr(Z init_kwargs isinstancestrlistrr'r:rr@ enumerateziprefinditerr-reversedr.Z_check_special_mm_tokensrIZ zeros_liker,itemsrLr)r2r6r7ZvideosZaudior3Z output_kwargsZ image_inputsZbatched_imagesZbatch_num_cropsZ batch_idxrGZ image_indexesnumidxZformatted_image_textrHZ text_inputsZ array_idsZmm_token_type_idsr)r?r2r__call__Osf  &   zGemma3Processor.__call__cO|jj|i|S)z This method forwards all its arguments to GemmaTokenizerFast's [`~PreTrainedTokenizer.batch_decode`]. Please refer to the docstring of this method for more information. )r( batch_decoder2argsr3rrrr`zGemma3Processor.batch_decodecOr_)z This method forwards all its arguments to GemmaTokenizerFast's [`~PreTrainedTokenizer.decode`]. Please refer to the docstring of this method for more information. )r(decoderarrrrdrczGemma3Processor.decodecCs(|jjdg}|jj}tt||S)NrP)r(model_input_namesr'rUdictfromkeys)r2Ztokenizer_input_namesZimage_processor_input_namesrrrresz!Gemma3Processor.model_input_names)Nr!)NNNN)rrr attributesZ valid_kwargsZimage_processor_classZtokenizer_classrr1rrrr rr rrr^r`rdpropertyre __classcell__rrr4rr&3s6 Ir&)rXtypingrrrnumpyrIZfeature_extraction_utilsrZ image_utilsrrZprocessing_utilsr r r r Ztokenization_utils_baser rutilsrrrr&__all__rrrrs   {