
    Uh'C                         d dl mZmZmZ d dlZd dlmZmZm	Z	m
Z
 d dlmZmZ ddlmZ ddlmZmZmZ ddlmZmZmZmZ  G d	 d
ed      Z G d ded      Z G d de	      ZdgZy)    )ListOptionalUnionN)ImagesKwargsProcessingKwargsProcessorMixinUnpack)PreTokenizedInput	TextInput   )BatchFeature)
ImageInputconcatenate_listmake_flat_list_of_images)
VideoInputVideoMetadata
load_videomake_batched_videosc                   >    e Zd ZU ee   ed<   ee   ed<   ee   ed<   y)InternVLImagesKwargscrop_to_patchesmin_patchesmax_patchesN)__name__
__module____qualname__r   bool__annotations__int     /var/www/catia.catastroantioquia-mas.com/valormas/lib/python3.12/site-packages/transformers/models/internvl/processing_internvl.pyr   r   &   s     d^###r!   r   F)totalc                   ,    e Zd ZU eed<   ddiddii dZy)InternVLProcessorKwargsimages_kwargspadding_sideleftr   T)text_kwargsr&   videos_kwargsN)r   r   r   r   r   	_defaultsr    r!   r"   r%   r%   ,   s,    '' F
 t
 Ir!   r%   c                       e Zd ZdZg dZddgZdZdZdZ	 	 	 	 	 d de	f fdZ
d	ee   d
ee	   dee	   dej                  dej                  dej                  fdZ	 	 	 	 d!dee   d	eeeeee   ee   f      dee   dee   def
dZ	 d"dedee	   deeee	f   fdZd Zd Z e!d        Z"	 	 d#deedf   dee	   dededejF                  f
dZ$ xZ%S )$InternVLProcessoraM  
    Constructs a InternVL processor which wraps a [`AutoImageProcessor`] and
    [`PretrainedTokenizerFast`] tokenizer into a single processor that inherits both the image processor and
    tokenizer functionalities. See the [`~InternVLProcessor.__call__`] and [`~InternVLProcessor.decode`] for more information.
    Args:
        image_processor ([`AutoImageProcessor`], *optional*):
            The image processor is a required input.
        tokenizer ([`PreTrainedTokenizer`, `PreTrainedTokenizerFast`], *optional*):
            The tokenizer is a required input.
        video_processor ([`AutoVideoProcessor`], *optional*):
            The video processor is a required input.
        image_seq_length (`int`, *optional*, defaults to 256):
            The number of image token to use per image patch. it should be set so that:
            image_seq_length = (config.image_size // config.patch_size) ** 2 * (config.scale_factor**2)
        chat_template (`str`, *optional*): A Jinja template which will be used to convert lists of messages
            in a chat into a tokenizable string.
    )image_processor	tokenizervideo_processorchat_templateimage_seq_lengthAutoImageProcessorAutoVideoProcessorAutoTokenizerc                     || _         |j                  | _        |j                  | _        |j                  | _        |j
                  | _        |j                  | _        t        | $  |||fd|i| y )Nr1   )
r2   start_image_tokenend_image_tokencontext_image_tokenimage_tokenvideo_tokencontext_image_token_idimage_token_idsuper__init__)selfr.   r/   r0   r2   r1   kwargs	__class__s          r"   r?   zInternVLProcessor.__init__U   sp     !1!*!<!<(88$88$00'>>)_lTaleklr!   textimage_num_patchesvideo_num_patchesimage_num_patches_indicesvideo_num_patches_indicesvideo_patch_indicesc	           	      R    d}	d}
g }g }g }|D ]  }|} j                   |v s j                  |v r j                   |v r؉ j                  |vs7|j                   j                         |j                   j                        k  r|	dkD  r||	dz
     nd}||	   }|j                  |||        |j	                   j                   dd      }|j                   j
                    j                    j                  z  ||	   z    j                          |	dz  }	n|
dkD  r||
dz
     nd}||
   }|
dkD  r||   nd}||dz
     }|j                  |||        t        |||       dj                   fdt        t                    D              }|j                  |       |j	                   j                  dd      }|
dz  }
 j                   |v r j                  |v rd|v r)|j                  d      }|j	                  d|d      }d|v r)|j                  |        |||	|
fS )z
        Processes interleaved text with <image> and <video> placeholders, replacing them with appropriate
        image and video tokens while keeping track of the patches used.
        r      z<placeholder>
c              3      K   | ]D  }d |dz    dj                    j                  j                  z  |   z   j                    F yw)FramerJ   z: N)r7   r:   r2   r8   ).0inum_patchesr@   s     r"   	<genexpr>z?InternVLProcessor._insert_media_placeholders.<locals>.<genexpr>   ss      -  Awb)?)?(@AQAQTXTiTiAilwxylzAz@{  }A  }Q  }Q  |R  S-s   A
A)r:   r;   indexappendreplacer7   r2   r8   listjoinrangelenpop)r@   rC   image_pixel_valuesvideo_pixel_valuesrD   rE   rF   rG   rH   image_indexvideo_indexprocessed_textimage_video_patchesreplace_stringsprompt
new_promptstart_index	end_indexcurrent_patch_indexend_patch_indexvideo_promptreplace_strrP   s   `                     @r"   _insert_media_placeholdersz,InternVLProcessor._insert_media_placeholdersg   s       &	.FJ""j0D4D4D
4R##z1$$J6!''(8(89J<L<LTM]M]<^^ Q\^_P_";K!O"LefK 9+ FI'../A+i/XY!+!3!3D4D4DoWX!YJ#**11243C3CdF[F[3[^op{^|3|2}  C  S  S  ~T  U  1$K
 S^`aRa*=kAo*Ngh'&9+&FOT_bcTc";<O"PijK 9/A:M NI'../A+i/XY"&'89L_']"^K#'99 -!&s;'7!8- $L $**<8!+!3!3D4D4DoWX!YJ1$KA ""j0D4D4D
4RB "Z/-11!4'//aP
 "Z/ !!*-M&	.P 2KLLr!   imagesvideosrA   returnc           
         |t        d       | j                  t        fd| j                  j                  i|}t        |t        t        f      s|g}g }g }i }	d}
d}t        j                  dg      }t        j                  dg      }t        j                  dg      }|Yt        |      } | j                  dd|i|d   }|j                  d      }|j                  d      }
t        j                  |      }|t        |      }|D cg c]  }t        |       }}t        j                  |      } | j                   dd	|i|d
   }|D cg c]  }t#        |      D ]  }d  }}}|j                  d      j%                  dd      }t        j                  |      }||`| j'                  ||
||||||      \  }}}}||t        |      k7  rt        d      ||t        |      k7  rt        d      dt)        |      i}	|d   j                  dd      } | j                  |fi |d   }| j+                  ||dg       t-        i ||	|      S c c}w c c}}w )a	  
        Main method to prepare for the model one or several sequences(s) and image(s). This method forwards the `text`
        and `kwargs` arguments to PreTrainedTokenizerFast's [`~PreTrainedTokenizerFast.__call__`] to encode the text if `text`
        is not `None`, otherwise encode default OCR queries which depends on the `format`, `box`, `color`, `multi_page` and
        `crop_to_patches` arguments. To prepare the vision inputs, this method forwards the `images` and `kwrags` arguments to
        GotOcr2ImageProcessor's [`~GotOcr2ImageProcessor.__call__`] if `images` is not `None`.

        Args:
            images (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`, `List[torch.Tensor]`):
                The image or batch of images to be prepared. Each image can be a PIL image, NumPy array or PyTorch
                tensor. Both channels-first and channels-last formats are supported.
            text (`str`, `List[str]`, `List[List[str]]`):
                The sequence or batch of sequences to be encoded. Each sequence can be a string or a list of strings
                (pretokenized string). If the sequences are provided as list of strings (pretokenized), you must set
                `is_split_into_words=True` (to lift the ambiguity with a batch of sequences).
            videos (`np.ndarray`, `torch.Tensor`, `List[np.ndarray]`, `List[torch.Tensor]`):
                The image or batch of videos to be prepared. Each video can be a 4D NumPy array or PyTorch
            return_tensors (`str` or [`~utils.TensorType`], *optional*):
                If set, will return tensors of a particular framework. Acceptable values are:
                - `'tf'`: Return TensorFlow `tf.constant` objects.
                - `'pt'`: Return PyTorch `torch.Tensor` objects.
                - `'np'`: Return NumPy `np.ndarray` objects.
                - `'jax'`: Return JAX `jnp.ndarray` objects.

        Returns:
            [`BatchFeature`]: A [`BatchFeature`] with the following fields:

            - **input_ids** -- List of token ids to be fed to a model. Returned when `text` is not `None`.
            - **attention_mask** -- List of indices specifying which tokens should be attended to by the model (when
              `return_attention_mask=True` or if *"attention_mask"* is in `self.model_input_names` and if `text` is not
              `None`).
            - **pixel_values** -- Pixel values to be fed to a model. Returned when `images` is not `None`.
        NzYou have to specify text.tokenizer_init_kwargsr   rj   r&   rP   pixel_valuesrk   r*   rJ   pixel_values_videoszONumber of image placeholders in the prompt does not match the number of images.zONumber of video placeholders in the prompt does not match the number of videos.r)   return_tensorsimage)
modalities)datatensor_typer    )
ValueError_merge_kwargsr%   r/   init_kwargs
isinstancerU   tuplenparrayr   r.   rY   cumsumr   rX   r0   rW   flattenri   r   _check_special_mm_tokensr   )r@   rj   rC   audiork   rA   output_kwargsrD   rE   image_videos_inputsrZ   r[   rF   rH   rG   image_inputsvideonum_frames_per_videovideo_inputsframes_r_   r\   r]   rq   text_inputss                             r"   __call__zInternVLProcessor.__call__   s   R <899***#
"&.."<"<
 
 $u.6D  !!$&HHaSM! hhsm$&HHaSM!-f5F/4//`v`A_`L , 0 0 ?!-!1!1.!A(*		2C(D%(0F<B#C5CJ#C #C"$)),@"A/4//`v`A_`L1E ]vuU[} ]! ] ] ]!-!1!12G!H!P!PQRTU!V(*		2C(D%!3BFBaBa""!!))#	C?D%{K !kS[&@ !rss!kS[&@ !rss $23CDW3X"Y&}599:JDQ$dnnTJ]=-IJ%%dKWI%N!GK!G3F!GUcdd= $D !^s   IImetadata
num_framesinitial_shiftc                     ||n|j                   }|du r|j                   |z  dz  }t        j                  ||j                   |j                   |z        j                  t              }|S )a  
        The function to generate indices of frames to sample from a video.

        Args:
            metadata (`VideoMetadata`):
                `VideoMetadata` object containing metadata about the video, such as "total_num_frames" or "fps".
            num_frames (`int`, *optional*):
                Number of frames to sample uniformly. If None, all frames are sampled.
            initial_shift (`bool`, `float` or `int`, defaults to `0`):
                The initial shift to apply when sampling frames. If `True`, the shift is set so that frames are sampled from the middle of the video.

        Returns:
            `np.ndarray`: Array of frame indices to sample.
        T   )total_num_framesr{   arangeastyper   )r@   r   r   r   indicess        r"   sample_indices_fnz#InternVLProcessor.sample_indices_fn  sl    " $.#9Zx?X?X
D $55
BQFM))M8+D+DhF_F_blFlmtt
 r!   c                 :     | j                   j                  |i |S )z
        This method forwards all its arguments to PreTrainedTokenizerFast's [`~PreTrainedTokenizer.batch_decode`]. Please
        refer to the docstring of this method for more information.
        )r/   batch_decoder@   argsrA   s      r"   r   zInternVLProcessor.batch_decode'  s     
 +t~~**D;F;;r!   c                 :     | j                   j                  |i |S )z
        This method forwards all its arguments to PreTrainedTokenizerFast's [`~PreTrainedTokenizer.decode`]. Please refer to
        the docstring of this method for more information.
        )r/   decoder   s      r"   r   zInternVLProcessor.decode.  s     
 %t~~$$d5f55r!   c                     | j                   j                  }| j                  j                  }t        |      t        |      z   S )N)r/   model_input_namesr.   rU   )r@   tokenizer_input_namesimage_processor_input_namess      r"   r   z#InternVLProcessor.model_input_names5  s;     $ @ @&*&:&:&L&L#)*T2M-NNNr!   r   r   backendc                 @      fd}t        |||      \  }}||fS )a  
        Loads `video` to a numpy array.

        Args:
            video (`str` or `VideoInput`):
                The video to convert to the numpy array format. Can be a link to video or local path.
            num_frames (`int`, *optional*):
                Number of frames to sample uniformly. If not passed, the whole video is loaded.
            backend (`str`, *optional*, defaults to `"pyav"`):
                The backend to use when loading the video. Can be any of ["decord", "pyav", "opencv", "torchvision"]. Defaults to "pyav".
            initial_shift (`bool`, *optional*, defaults to `True`):
                The initial shift to apply when sampling frames. If `True`, the shift is set so that frames are sampled from the middle of the video.

        Returns:
            Tuple[`np.array`, Dict]: A tuple containing:
                - Numpy array of frames in RGB (shape: [num_frames, height, width, 3]).
                - Metadata dictionary.
        c                 0     j                   | fd|S )N)r   r   )r   )r   	fn_kwargsr   r   r@   s     r"   sample_indices_fn_funczGInternVLProcessor._load_video_for_model.<locals>.sample_indices_fn_funcW  s$    )4))(tzYftjsttr!   )r   r   )r   )r@   r   r   r   r   rA   r   r   s   ` ` `   r"   _load_video_for_modelz'InternVLProcessor._load_video_for_model<  s(    6	u %UGOefxhr!   )NNN   N)NNNN)NT)pyavT)&r   r   r   __doc__
attributesvalid_kwargsimage_processor_classvideo_processor_classtokenizer_classr   r?   rU   strr{   ndarrayri   r   r   r   r   r
   r   r   r	   r%   r   r   r   r   floatr   r   r   propertyr   r|   r   __classcell__)rB   s   @r"   r-   r-   9   s   $ EJL 10%O  #m
 m$>M3i>M
  9>M  9>M $&::>M $&::>M  ZZ>MD (,hl'+de$de uY(94	?DQbLccdede
 $de 01de 
deN sw%3;C=X]^bdikn^nXo4<6 O O "S,&' SM 	
  
r!   r-   )typingr   r   r   numpyr{   transformers.processing_utilsr   r   r   r	   $transformers.tokenization_utils_baser
   r   image_processing_utilsr   image_utilsr   r   r   video_utilsr   r   r   r   r   r%   r-   __all__r    r!   r"   <module>r      sl   " ) (   N 2 
 V U<u 
.e 
b bJ	 
r!   