o ۷iM(@sddlmZmZmZddlmZddlmZmZm Z m Z m Z m Z ddl mZmZe r7ddlmZddlmZer@dd lmZe rMdd lZdd lmZe eZeed d d GdddeZd S))AnyUnionoverload)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)$MODEL_FOR_VISION_2_SEQ_MAPPING_NAMEST) has_tokenizerhas_image_processorc seZdZdZdZdZdZdZdZe ddZ fddZ dd d Z e d eed fd edeeeeffddZe d eeeed fd edeeeeeffddZd eeeed ed fffdd ZdddZddZddZZS)ImageToTextPipelinea Image To Text pipeline using a `AutoModelForVision2Seq`. This pipeline predicts a caption for a given image. Unless the model you're using explicitly sets these generation parameters in its configuration files (`generation_config.json`), the following default values will be used: - max_new_tokens: 256 Example: ```python >>> from transformers import pipeline >>> captioner = pipeline(model="ydshieh/vit-gpt2-coco-en") >>> captioner("https://huggingface.co/datasets/Narsil/image_dummy/raw/main/parrots.png") [{'generated_text': 'two birds are standing next to each other '}] ``` Learn more about the basics of using a pipeline in the [pipeline tutorial](../pipeline_tutorial) This image to text pipeline can currently be loaded from pipeline() using the following task identifier: "image-to-text". See the list of available models on [huggingface.co/models](https://huggingface.co/models?pipeline_tag=image-to-text). TF)max_new_tokenscs>tj|i|t|d||jdkrtdStdS)Nvisiontf)super__init__r check_model_type frameworkrr)selfargskwargs __class__Z/home/ubuntu/vllm_env/lib/python3.10/site-packages/transformers/pipelines/image_to_text.pyrTs  zImageToTextPipeline.__init__NcCsi}i}|dur ||d<|dur||d<|dur||d<|dur1|dur,d|vr,td|||jdur;|j|d<|jdurJ|j|d<|j|d<||ifS)Nprompttimeoutrzp`max_new_tokens` is defined both as an argument and inside `generate_kwargs` argument, please use only 1 versionassistant_model tokenizerassistant_tokenizer) ValueErrorupdater(r*r))rrgenerate_kwargsr&r'forward_paramspreprocess_paramsr$r$r%_sanitize_parameters[s(       z(ImageToTextPipeline._sanitize_parametersinputsz Image.Imager!returncKdSNr$rr1r!r$r$r%__call__vzImageToTextPipeline.__call__cKr3r4r$r5r$r$r%r6yr7c s6d|vr |d}|durtdtj|fi|S)a Assign labels to the image(s) passed as inputs. Args: inputs (`str`, `list[str]`, `PIL.Image` or `list[PIL.Image]`): The pipeline handles three types of images: - A string containing a HTTP(s) link pointing to an image - A string containing a local path to an image - An image loaded in PIL directly The pipeline accepts either a single image or a batch of images. max_new_tokens (`int`, *optional*): The amount of maximum tokens to generate. By default it will use `generate` default. generate_kwargs (`Dict`, *optional*): Pass it to send all of these arguments directly to `generate` allowing full control of this function. timeout (`float`, *optional*, defaults to None): The maximum time in seconds to wait for fetching images from the web. If None, no timeout is set and the call may block forever. Return: A list or a list of list of `dict`: Each result comes as a dictionary with the following key: - **generated_text** (`str`) -- The generated text. imagesNzBCannot call the image-to-text pipeline without an inputs argument!)popr+rr6r5r"r$r%r6|s  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)r8return_tensorsptF)textadd_special_tokensr input_ids pix2struct)r8 header_textr;zvision-encoder-decoder)r;z Model type z- does not support conditional text generation)rlogger warning_once isinstancestrr+typemodelconfig model_typeimage_processorrtodtyper)r? cls_token_idtorchtensor unsqueezer,)rimager&r'rI model_inputsr? text_inputsr$r$r% preprocesssH            zImageToTextPipeline.preprocesscKspd|vrt|dtrtdd|dDrd|d<d|vr#|j|d<||jj}|jj|fi||}|S)Nr?css|]}|duVqdSr4r$).0xr$r$r% sz/ImageToTextPipeline._forward..generation_config)rDlistallrXr9rGmain_input_namegenerate)rrRr-r1 model_outputsr$r$r%_forwards  zImageToTextPipeline._forwardcCs0g}|D]}d|jj|ddi}||q|S)Ngenerated_textT)skip_special_tokens)r)decodeappend)rr]records output_idsrecordr$r$r% postprocesss zImageToTextPipeline.postprocess)NNNN)NN)__name__ __module__ __qualname____doc___pipeline_calls_generate_load_processor_load_image_processor_load_feature_extractor_load_tokenizerr_default_generation_configrr0rrrErrYdictr6rTr^rf __classcell__r$r$r"r%r.s(  ,8& $3r)typingrrr generationrutilsrrr r r r baserrPILr image_utilsrmodels.auto.modeling_tf_autorrNmodels.auto.modeling_autor get_loggerrgrBrr$r$r$r%s