sg] DdZddlZddlmZmZmZmZmZmZm Z ddl Z ddl m Z mZmZddlmZmZddlmZmZmZmZmZmZmZmZmZmZmZddlm Z m!Z!m"Z"m#Z#m$Z$m%Z%dd l&m'Z'e%jPe)Z*e$rddl+Z+Gd d eZ,d ed eee jZfdZ.ded efdZ/dee0e0fdee0e0fd e0fdZ1 ddede e0ee0e0fee0ee0fde e0ee0e0fee0ee0fdee e2efd e3f dZ4de e2e fd eeeffdZ5de e2e fd efdZ6Gdde Z7y)z"Image processor class for Pixtral.N)AnyCallableDictListOptionalTupleUnion)BaseImageProcessor BatchFeature get_size_dict)resizeto_channel_dimension_format) ChannelDimension ImageInputPILImageResamplingget_image_sizeinfer_channel_dimension_formatis_scaled_imageis_valid_imageto_numpy_array valid_imagesvalidate_kwargsvalidate_preprocess_arguments) TensorTypeis_torch_deviceis_torch_dtypeis_torch_tensoris_vision_availablelogging)requires_backendsceZdZddZy)BatchMixFeaturec t|dgddl}i}|jd}|et|dkDrW|d}t |rnFt |t st|st |tr|}ntdt |d|jD]\}}t |tr9|D cgc]'} | D] } t| s| j|i|")c} } ||<Ot ||jr'|j|r|j|i|||<t ||jr||j|||<|||<||_|Scc} } w)a  Send all values to device by calling `v.to(*args, **kwargs)` (PyTorch only). This should support casting in different `dtypes` and sending the `BatchFeature` to a different `device`. Args: args (`Tuple`): Will be passed to the `to(...)` function of the tensors. kwargs (`Dict`, *optional*): Will be passed to the `to(...)` function of the tensors. Returns: [`BatchFeature`]: The same instance after modification. torchrNdevicez*Attempting to cast a BatchFeature to type z. This is not supported.)r&)r!r%getlenr isinstancestrrint ValueErroritemslistrtoTensoris_floating_pointdata) selfargskwargsr%new_datar&argkvsampleelements g/var/www/html/venv/lib/python3.12/site-packages/transformers/models/pixtral/image_processing_pixtral.pyr/zBatchMixFeature.to4sf $ *H% >c$i!mq'Cc"C%)=CQTAU!#McRUhZWo!pqqJJL DAq!T">?4:vDKYhipYqJGJJ/// Au||,1H1H1K"addD3F3 Au||,1Cdd&d1     s 7E& E&N)returnr#)__name__ __module__ __qualname__r/r<r#r#3s.rBr#imagesr=clt|r|gg}|St|ttfr!t |dkDrt|dr|g}|St|ttfrLt |dkDr>t|dttfr%t |ddkDrt|ddr |St d)z Convert a single image or a list of images to a list of numpy arrays. Args: images (`ImageInput`): A single image or a list of images. Returns: A list of numpy arrays. rz]Invalid input type. Must be a single image, a list of images, or a list of batches of images.)rr)r.tupler(r,)rCs r<make_list_of_imagesrFfsf(" M FT5M *s6{Q>RXYZR[C\ M 6D%=) K!O vay4- 0 q NQ  6!9Q< ( M k  rBimagecVttdgt|tjjs|S|j dk(r|S|j d}tjjd|jd}|j|d||j d}|S)z Converts an image to RGB format. Only converts if the image is of type PIL.Image.Image, otherwise returns the image as is. Args: image (Image): The image to convert. visionRGBRGBAWHITE)rr) r!convert_to_rgbr)PILImagemodeconvertnewsizepaste)rG new_images r<rMrMsnxj1 eSYY__ -  zzU  MM& !E fejj':I OOE65)!!%(I rB image_size patch_sizec~|\}}t|ttfr|n||f\}}|dz |zdz}|dz |zdz}||fS)aP Calculate the number of image tokens given the image size and patch size. Args: image_size (`Tuple[int, int]`): The size of the image as `(height, width)`. patch_size (`Tuple[int, int]`): The patch size as `(height, width)`. Returns: `int`: The number of image tokens. )r)rEr.)rVrWheightwidth patch_height patch_widthnum_width_tokensnum_height_tokenss r<_num_image_tokensr`saMFE.8eT].S ZdfpYqL+ k1A5! 4q8 . ..rB input_imagerSinput_data_formatct|ttfr|n||f\}}t|ttfr|n||f\}}t||\}} t ||z | |z } | dkDrBt t j|| z }t t j| | z } t|| f||f\} } | |z| |zfS)aN Find the target (height, width) dimension of the output image after resizing given the input image and the desired size. Args: input_image (`ImageInput`): The image to resize. size (`int` or `Tuple[int, int]`): Max image size an input image can be. Must be a dictionary with the key "longest_edge". patch_size (`int` or `Tuple[int, int]`): The patch_size as `(height, width)` to use for resizing the image. If patch_size is an integer, `(patch_size, patch_size)` will be used input_data_format (`ChannelDimension`, *optional*): The channel dimension format of the input image. If unset, will use the inferred format from the input. Returns: `tuple`: The target (height, width) dimension of the output image after resizing. rY) r)rEr.rmaxr+mathceilr`) rarSrWrb max_height max_widthr\r]rZr[ratior_r^s r<get_resize_output_image_sizerjs0%/teT]$CD$PTJ .8eT].S ZdfpYqL+";0ABMFE #UY%6 7E qyTYYv~./DIIeem,-*;VUOl\gMh*i'' | +-= -K KKrB tensor_typec4tj|SN)r _get_is_as_tensor_fns)rks r<rnrns > / / <*Dg&/&;Ae,& "rBrGrrbc d|vr |d|df}nd|vrd|vr |d|df}n tdd|vrd|vr |d|df}n tdt||||}t|f||||d|S)a Resize an image. The shortest edge of the image is resized to size["shortest_edge"], with the longest edge resized to keep the input aspect ratio. Args: image (`np.ndarray`): Image to resize. size (`Dict[str, int]`): Dict containing the longest possible edge of the image. patch_size (`Dict[str, int]`): Patch size used to calculate the size of the output image. resample (`PILImageResampling`, *optional*, defaults to `PILImageResampling.BICUBIC`): Resampling filter to use when resiizing the image. data_format (`str` or `ChannelDimension`, *optional*): The channel dimension format of the image. If not provided, it will be the same as the input image. input_data_format (`ChannelDimension` or `str`, *optional*): The channel dimension format of the input image. If not provided, it will be inferred. rrZr[z@size must contain either 'longest_edge' or 'height' and 'width'.zGpatch_size must contain either 'shortest_edge' or 'height' and 'width'.)rSrWrb)rSrxrrb)r,rjr) r3rGrSrWrxrrbr5 output_sizes r<rzPixtralImageProcessor.resize;s8 T !($~*>?D  'T/NDM2D_` ` z !g&;$X. 70CDJfg g2 !/    #/     rBrCrc ||n |j}t|d}||n |j}||n |j}||n |j}||n |j }||n |j }||n |j}| | n |j} | | n |j} | | n |j} t|j|jt|}t|ds t!dt#|||| | |||| r(|Dcgc]}|Dcgc] }t%|c}}}}|Dcgc]}|Dcgc] }t'|c}}}}t)|ddr|rt*j-d|t/|dd}g}g}|D]}g}g}|D]s}|r|j1||||| }|r|j3||| }|r|j5|| | | }|j7||j7t9||u|j7||j7||Dcgc]}|Dcgc]}t;|| | c} }}}|Dcgc]}|Dcgc]}t=|| c}}}}t?||d dScc}wcc}}wcc}wcc}}wcc}wcc}}wcc}wcc}}w)a Preprocess an image or batch of images. Args: images (`ImageInput`): Image to preprocess. Expects a single or batch of images with pixel values ranging from 0 to 255. If passing in images with pixel values between 0 and 1, set `do_rescale=False`. do_resize (`bool`, *optional*, defaults to `self.do_resize`): Whether to resize the image. size (`Dict[str, int]`, *optional*, defaults to `self.size`): Describes the maximum input dimensions to the model. patch_size (`Dict[str, int]`, *optional*, defaults to `self.patch_size`): Patch size in the model. Used to calculate the image after resizing. resample (`int`, *optional*, defaults to `self.resample`): Resampling filter to use if resizing the image. This can be one of the enum `PILImageResampling`. Only has an effect if `do_resize` is set to `True`. do_rescale (`bool`, *optional*, defaults to `self.do_rescale`): Whether to rescale the image. rescale_factor (`float`, *optional*, defaults to `self.rescale_factor`): Rescale factor to rescale the image by if `do_rescale` is set to `True`. do_normalize (`bool`, *optional*, defaults to `self.do_normalize`): Whether to normalize the image. image_mean (`float` or `List[float]`, *optional*, defaults to `self.image_mean`): Image mean to use for normalization. Only has an effect if `do_normalize` is set to `True`. image_std (`float` or `List[float]`, *optional*, defaults to `self.image_std`): Image standard deviation to use for normalization. Only has an effect if `do_normalize` is set to `True`. do_convert_rgb (`bool`, *optional*, defaults to `self.do_convert_rgb`): Whether to convert the image to RGB. return_tensors (`str` or `TensorType`, *optional*): The type of tensors to return. Can be one of: - Unset: Return a list of `np.ndarray`. - `TensorType.TENSORFLOW` or `'tf'`: Return a batch of type `tf.Tensor`. - `TensorType.PYTORCH` or `'pt'`: Return a batch of type `torch.Tensor`. - `TensorType.NUMPY` or `'np'`: Return a batch of type `np.ndarray`. - `TensorType.JAX` or `'jax'`: Return a batch of type `jax.numpy.ndarray`. data_format (`ChannelDimension` or `str`, *optional*, defaults to `ChannelDimension.FIRST`): The channel dimension format for the output image. Can be one of: - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format. - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format. - Unset: Use the channel dimension format of the input image. input_data_format (`ChannelDimension` or `str`, *optional*): The channel dimension format for the input image. If unset, the channel dimension format is inferred from the input image. Can be one of: - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format. - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format. - `"none"` or `ChannelDimension.NONE`: image in (height, width) format. NTr)captured_kwargsvalid_processor_keysrzkInvalid image type. Must be of type PIL.Image.Image, numpy.ndarray, torch.Tensor, tf.Tensor or jax.ndarray.)ryrzr{r|r}rwrSrxzIt looks like you are trying to rescale already rescaled images. If the input images have pixel values between 0 and 1, set `do_rescale=False` to avoid rescaling them again.)rGrSrWrxrb)rGscalerb)rGmeanstdrb)input_channel_dim)rv image_sizes)r2rk) rWr rwrSrxryrzr{r|r}r~rkeysrrFrr,rrMrrlogger warning_oncerrrescale normalizeappendrrrsr#)r3rCrwrSrWrxryrzr{r|r}r~rrrbr5 images_listrG batch_imagesbatch_image_sizes sample_imagesrs r< preprocessz PixtralImageProcessor.preprocessrsD$.#9Zt ":F !*!6IDNN 'tTYY'38#-#9Zt +9+E4K^K^'3'?|TEVEV #-#9Zt !*!6IDNN +9+E4K^K^ DLfLfg)&1 KN+:  &!)%!  U`a6vFeN51FaKaR]]v6B%u-B] ] ;q>!, -*   s   $ >{1~a?P Q  ( 2MFK& M KK#!#-!)*; (E LLuN^oLpE NN#*)Wh+E e$"">%9J#KL' M(    '  $ $[ 11 28' pv vfk ( O` a v  eppZ`fUU)%@Up p[Qb$cquvveGaC]R w Vps` J)J$ ,J)9 J4J/ J4 J?J: #J?0 K 9K K $J)/J4:J?K )r>r?r@__doc__model_input_namesrBICUBICboolrr*r+r floatrrrnpndarrayrrFIRSTrrrNrOr __classcell__)rs@r<rurusB((#%)'9'A'A,3!:>9=#, , 38n, cN , % ,  , c5j), , U5$u+#567, E%e"456, ,  , f(:'A'A>BDH5 zz5 38n5 cN 5 % 5 eC)9$9:; 5 $E#/?*?$@A5  5 t#%)'+ $!:>9=#;?2B2H2HDHWwWwWw38n Ww cN Ww % WwWwWwWwU5$u+#567WwE%e"456WwWw!sJ!78Ww./Ww$E#/?*?$@AWw" #WwrBrurm)8rretypingrrrrrrr numpyrimage_processing_utilsr r r image_transformsrr image_utilsrrrrrrrrrrrutilsrrrrrr utils.import_utilsr! get_loggerr>rrNr#rrFrMr+r`r*rErjrnrsrurArBr<rs) DDDUU    po3   H %/l/f tD4D/ED*4/%S//uS#X/SV/0AE $L$L U38_d3is; <$Lc5c?DIuSzAB$L c+;&; <= $L  $LP=uS*_'==%RZHZB[=%Z*@S`w.`wrB