o Zh&@sddlmZmZddlmZddlmZmZmZm Z m Z m Z ddl m Z mZe r5ddlmZddlmZer>dd lmZerKdd lZdd lmZe eZeed d d Gddde Zd S))ListUnion)GenerationConfig)add_end_docstringsis_tf_availableis_torch_availableis_vision_availableloggingrequires_backends)Pipelinebuild_pipeline_init_args)Image) load_image)'TF_MODEL_FOR_VISION_2_SEQ_MAPPING_NAMESN)$MODEL_FOR_VISION_2_SEQ_MAPPING_NAMEST)Z has_tokenizerZhas_image_processorcszeZdZdZdZeddZfddZddd Zdd e e e e d e d fffd d Z dddZ ddZddZZS)ImageToTextPipelinea Image To Text pipeline using a `AutoModelForVision2Seq`. This pipeline predicts a caption for a given image. Unless the model you're using explicitly sets these generation parameters in its configuration files (`generation_config.json`), the following default values will be used: - max_new_tokens: 256 Example: ```python >>> from transformers import pipeline >>> captioner = pipeline(model="ydshieh/vit-gpt2-coco-en") >>> captioner("https://huggingface.co/datasets/Narsil/image_dummy/raw/main/parrots.png") [{'generated_text': 'two birds are standing next to each other '}] ``` Learn more about the basics of using a pipeline in the [pipeline tutorial](../pipeline_tutorial) This image to text pipeline can currently be loaded from pipeline() using the following task identifier: "image-to-text". See the list of available models on [huggingface.co/models](https://huggingface.co/models?pipeline_tag=image-to-text). T)max_new_tokenscs>tj|i|t|d||jdkrtdStdS)NZvisiontf)super__init__r Zcheck_model_type frameworkrr)selfargskwargs __class__S/var/www/auris/lib/python3.10/site-packages/transformers/pipelines/image_to_text.pyrPs  zImageToTextPipeline.__init__NcCsi}i}|dur ||d<|dur||d<|dur||d<|dur1|dur,d|vr,td|||jdur;|j|d<|jdurJ|j|d<|j|d<||ifS)Nprompttimeoutrzp`max_new_tokens` is defined both as an argument and inside `generate_kwargs` argument, please use only 1 versionassistant_model tokenizerassistant_tokenizer) ValueErrorupdater#r%r$)rrgenerate_kwargsr!r"Zforward_paramsZpreprocess_paramsrrr _sanitize_parametersWs(       z(ImageToTextPipeline._sanitize_parametersinputsz Image.Imagec s6d|vr |d}|durtdtj|fi|S)a Assign labels to the image(s) passed as inputs. Args: inputs (`str`, `List[str]`, `PIL.Image` or `List[PIL.Image]`): The pipeline handles three types of images: - A string containing a HTTP(s) link pointing to an image - A string containing a local path to an image - An image loaded in PIL directly The pipeline accepts either a single image or a batch of images. max_new_tokens (`int`, *optional*): The amount of maximum tokens to generate. By default it will use `generate` default. generate_kwargs (`Dict`, *optional*): Pass it to send all of these arguments directly to `generate` allowing full control of this function. timeout (`float`, *optional*, defaults to None): The maximum time in seconds to wait for fetching images from the web. If None, no timeout is set and the call may block forever. Return: A list or a list of list of `dict`: Each result comes as a dictionary with the following key: - **generated_text** (`str`) -- The generated text. imagesNzBCannot call the image-to-text pipeline without an inputs argument!)popr&r__call__)rr*rrrr r-rs  zImageToTextPipeline.__call__cCst||d}|durtdt|tstdt|d|jjj }|dkrY|j ||j d}|j dkr:| |j }|j|dd j}|jjg|}t|d }|d |inY|d krr|j |||j d }|j dkrq| |j }n@|dkr|j ||j d}|j dkr| |j }|j||j d}||ntd|d|j ||j d}|j dkr| |j }|jjj dkr|durd|d <|S)N)r"uPassing `prompt` to the `image-to-text` pipeline is deprecated and will be removed in version 4.48 of 🤗 Transformers. Use the `image-text-to-text` pipeline insteadz&Received an invalid text input, got - zy - but expected a single string. Note also that one single text can be provided for conditional image to text generation.git)r+return_tensorsptF)textZadd_special_tokensr input_idsZ pix2struct)r+Z header_textr/zvision-encoder-decoder)r/z Model type z- does not support conditional text generation)rloggerZ warning_once isinstancestrr&typemodelconfig model_typeZimage_processorrtoZ torch_dtyper$r2Z cls_token_idtorchZtensorZ unsqueezer')rimager!r"r9 model_inputsr2Z text_inputsrrr preprocesssH            zImageToTextPipeline.preprocesscKspd|vrt|dtrtdd|dDrd|d<d|vr#|j|d<||jj}|jj|fi||}|S)Nr2css|]}|duVqdSNr).0xrrr sz/ImageToTextPipeline._forward..generation_config)r4listallrCr,r7Zmain_input_namegenerate)rr=r(r* model_outputsrrr _forwards  zImageToTextPipeline._forwardcCs0g}|D]}d|jj|ddi}||q|S)NZgenerated_textT)Zskip_special_tokens)r$decodeappend)rrGrecordsZ output_idsrecordrrr postprocesss zImageToTextPipeline.postprocess)NNNNr?)NN)__name__ __module__ __qualname____doc__Z_pipeline_calls_generaterZ_default_generation_configrr)rr5rr-r>rHrM __classcell__rrrr r.s  ( $3r)typingrrZ generationrutilsrrrr r r baser rZPILrZ image_utilsrZmodels.auto.modeling_tf_autorr;Zmodels.auto.modeling_autorZ get_loggerrNr3rrrrr s