`i`ddlZddlZddlmZddlmZddlmZddlm Z m Z ddl Z ddl Z ddlmZmZmZmZmZmZmZmZmZmZmZddlmZmZmZmZmZm Z erqddl!Z"ddl#Z"e"j$j%Z&erQdd l'm(Z(e&j)e(j*e&j+e(j+e&j,e(j,e&j-e(j-e&j.e(j.e&j/e(j/iZ0niZ0erddl1Z1ej2e3Z4e d e j5d e6d e6e j5e6d fZ7Gd d eZ8GddeZ9GddeZ:e;edZ?GddeZ@dZAdZBde6fdZCdZDdZEdZFde j5deGfdZHdUd e=de6e7fd!ZI dUde e6e7e7fd e=de7fd"ZJ dUde e6e7e7fd e=de6e7fd#ZKde j5fd$ZL dVde j5d%e e e=eMe=d&ffde8fd'ZN dVde j5d(e e e8ee eGd?e e e;eeZdZejjZejjZdS)AnnotionFormatN)rr r!r(r+valuer,r$r%r&r.r.[s$%4:N$28MMMr%r.c\tot|tjjSN)r isinstancePILImageimgs r& is_pil_imager7cs!  EZSY_%E%EEr%c"eZdZdZdZdZdZdZdS) ImageTypepillowtorchnumpy tensorflowjaxN)rr r!r3TORCHNUMPY TENSORFLOWJAXr$r%r&r9r9gs' C E EJ CCCr%r9cNt|r tjSt|r tjSt |r tjSt|r tjSt|r tj Stdt|)NzUnrecognized image type ) r7r9r3rr?r r@r rAr rB ValueErrortypeimages r&get_image_typerHosE}ueE$##U} =U == > >>r%ct|p;t|p,t|pt|pt |Sr1)r7r rr r r5s r&is_valid_imagerJ}sG    vs 3 3 vs7K7K v|\_O`O` vdqrudvdvvr%imagesc8|otd|DS)Nc34K|]}t|VdSr1)rJ.0rGs r& z*is_valid_list_of_images..s*DDE.//DDDDDDr%all)rKs r&is_valid_list_of_imagesrSs$  DcDDVDDDDDDr%c,t|dtr d|DSt|dtjrtj|dSt|dt jrt j|dSdS)Nrcg|] }|D]}| Sr$r$)rOsublistitems r& z$concatenate_list..s%CCC7CC4CCCCr%axis)dim)r2listnpndarray concatenater;Tensorcat) input_lists r&concatenate_listrcs*Q-&&,CCJCCCC JqM2: . .,~jq1111 JqM5< 0 0,y++++,,r%ct|ttfr|D]}t|sdSnt |sdSdS)NFT)r2r\tuple valid_imagesrJ)imgsr6s r&rfrfsd$u &&  C$$ uu  D ! !u 4r%cht|ttfrt|dSdS)NrF)r2r\rerJr5s r& is_batchedris/#e}%%&c!f%%% 5r%rGreturnc|jtjkrdStj|dkotj|dkS)zV Checks to see whether the pixel values have already been rescaled to [0, 1]. Frr)dtyper]uint8minmaxrFs r&is_scaled_imagerps? {bhu 6%==A  4"&--1"44r%expected_ndimsc Ft|r|St|r|gSt|rP|j|dzkrt |}n0|j|kr|g}n!t d|dzd|d|jd|St dt |d)a Ensure that the output is a list of images. If the input is a single image, it is converted to a list of length 1. If the input is a batch of images, it is converted to a list of images. Args: images (`ImageInput`): Image of images to turn into a list of images. expected_ndims (`int`, *optional*, defaults to 3): Expected number of dimensions for a single input image. If the input image has a different number of dimensions, an error is raised. rz%Invalid image shape. Expected either z or z dimensions, but got z dimensions.ztInvalid image type. Expected either PIL.Image.Image, numpy.ndarray, torch.Tensor, tf.Tensor or jax.ndarray, but got .)rir7rJndimr\rDrErKrrs r&make_list_of_imagesrws& Fxf  ;.1, , ,&\\FF [N * *XFF.8J..P^..K...   0 $V  0 0 0  r%cRt|ttfr>td|Dr%td|Dr d|DSt|ttfrWt |rHt |ds|dj|kr|S|dj|dzkr d|DSt|r:t |s |j|kr|gS|j|dzkrt|Std|)a Ensure that the output is a flat list of images. If the input is a single image, it is converted to a list of length 1. If the input is a nested list of images, it is converted to a flat list of images. Args: images (`Union[list[ImageInput], ImageInput]`): The input image. expected_ndims (`int`, *optional*, defaults to 3): The expected number of dimensions for a single input image. Returns: list: A list of images or a 4d array of images. c3NK|] }t|ttfV!dSr1r2r\rerOimages_is r&rPz+make_flat_list_of_images..0KK 8dE]33KKKKKKr%c3:K|]}t|p| VdSr1rSr{s r&rPz+make_flat_list_of_images..2YYh'11A\YYYYYYr%cg|] }|D]}| Sr$r$rOimg_listr6s r&rXz,make_flat_list_of_images..s%???h??s????r%rrcg|] }|D]}| Sr$r$rs r&rXz,make_flat_list_of_images..s%CCCH(CC3CCCCCr%z*Could not make a flat list of images from r2r\rerRrSr7rurJrDrvs r&make_flat_list_of_imagesrsX" 6D%=))@ KKFKKK K K@ YYRXYYY Y Y@ @?F????&4-((D-DV-L-LD q " " fQin&F&FM !9>^a/ / /CCCCC Cf    6;.#@#@8O ;.1, , ,<<  J&JJ K KKr%c>t|ttfr4td|Drtd|Dr|St|ttfrXt |rIt |ds|dj|kr|gS|dj|dzkr d|DSt|r. r}r%c3:K|]}t|p| VdSr1rr{s r&rPz-make_nested_list_of_images.. rr%rrc,g|]}t|Sr$)r\rNs r&rXz.make_nested_list_of_images..s444EDKK444r%z]Invalid input type. Must be a single image, a list of images, or a list of batches of images.rrvs r&make_nested_list_of_imagesrsF 6D%=)) KKFKKK K K YYRXYYY Y Y  &4-((5-DV-L-L5 q " " fQin&F&F8O !9>^a/ / /44V444 4f"    6;.#@#@H:  ;.1, , ,LL> ! t u uur%ct|stdt|tr3t |t jjrtj|St|S)NzInvalid image type: ) rJrDrErr2r3r4r]arrayrr5s r&to_numpy_arrayrsl #  =;S ;;<<<C!A!Ax}} C==r% num_channels.c||nd}t|tr|fn|}|jdkrd\}}n9|jdkrd\}}n(|jdkrd\}}ntd|j|j||vr>|j||vr/t d |jd tjS|j||vr tjS|j||vr tj Std ) a[ Infers the channel dimension format of `image`. Args: image (`np.ndarray`): The image to infer the channel dimension of. num_channels (`int` or `tuple[int, ...]`, *optional*, defaults to `(1, 3)`): The number of channels of the image. Returns: The channel dimension of the image. Nrrqrq)r)rrz(Unsupported number of image dimensions: z4The channel dimension is ambiguous. Got image shape z. Assuming channels are the first dimension. Use the [input_data_format](https://huggingface.co/docs/transformers/main/internal/image_processing_utils#transformers.image_transforms.rescale.input_data_format) parameter to assign the channel dimension.z(Unable to infer channel dimension format) r2intrurDshapeloggerwarningrr"r#)rGr first_dimlast_dims r&infer_channel_dimension_formatr(s/$0#;<<L&0s&C&CUL??L zQ" 88 q" 88 q" 88PEJPPQQQ {9--%+h2G<2W2W K5; K K K    %% Y < / /%% X , . .$$ ? @ @@r%input_data_formatc|t|}|tjkr |jdz S|tjkr |jdz St d|)a Returns the channel dimension axis of the image. Args: image (`np.ndarray`): The image to get the channel dimension axis of. input_data_format (`ChannelDimension` or `str`, *optional*): The channel dimension format of the image. If `None`, will infer the channel dimension from the image. Returns: The channel dimension axis of the image. NrqrUnsupported data format: )rrr"rur#rD)rGrs r&get_channel_dimension_axisrOsf :5AA,222zA~ .3 3 3zA~ D1BDD E EEr% channel_dimc|t|}|tjkr|jd|jdfS|tjkr|jd|jdfSt d|)a Returns the (height, width) dimensions of the image. Args: image (`np.ndarray`): The image to get the dimensions of. channel_dim (`ChannelDimension`, *optional*): Which dimension the channel dimension is in. If `None`, will infer the channel dimension from the image. Returns: A tuple of the image's height and width. Nr)rrr"rr#rD)rGrs r&get_image_sizergsz4U;; &,,,{2 B// (- - -{2 B//B[BBCCCr% image_size max_height max_widthc|\}}||z }||z }t||}t||z}t||z} || fS)a Computes the output image size given the input image and the maximum allowed height and width. Keep aspect ratio. Important, even if image_height < max_height and image_width < max_width, the image will be resized to at least one of the edges be equal to max_height or max_width. For example: - input_size: (100, 200), max_height: 50, max_width: 50 -> output_size: (25, 50) - input_size: (100, 200), max_height: 200, max_width: 500 -> output_size: (200, 400) Args: image_size (`tuple[int, int]`): The image to resize. max_height (`int`): The maximum allowed height. max_width (`int`): The maximum allowed width. )rnr) rrrheightwidth height_scale width_scale min_scale new_height new_widths r&#get_image_size_for_max_height_widthrs_,MFE&Le#KL+..IVi'((JEI%&&I y  r% annotationct|trfd|vrbd|vr^t|dttfrWXY>Z\`3a3a /t 5r%ct|trjd|vrfd|vrbd|vr^t|dttfr.s+NN31#66NNNNNNr%rQrs r& valid_coco_detection_annotationsrs NN+NNN N NNr%c4td|DS)Nc34K|]}t|VdSr1)rrs r&rPz2valid_coco_panoptic_annotations..s+MM#055MMMMMMr%rQrs r&valid_coco_panoptic_annotationsrs MMMMM M MMr%timeoutcttdgt|tr[|ds|drGt jttj ||j }ntj |r t j|}n|dr|dd} t!j|}t jt|}nU#t&$r}t)d|d |d }~wwxYwt|t jjst+d t j|}|d }|S) a3 Loads `image` to a PIL Image. Args: image (`str` or `PIL.Image.Image`): The image to convert to the PIL Image format. timeout (`float`, *optional*): The timeout value in seconds for the URL request. Returns: `PIL.Image.Image`: A PIL Image. visionzhttp://zhttps://rz data:image/,rzIncorrect image source. Must be a valid URL starting with `http://` or `https://`, a valid path to an image file, or a base64 encoded string. Got z. Failed with NzuIncorrect format used for image. Should be an url linking to an image, a base64 string, a local path, or a PIL image.RGB)r load_imager2str startswithr3r4openrrequestsgetcontentospathisfilesplitbase64 decodebytesencode ExceptionrD TypeErrorImageOpsexif_transposeconvert)rGrb64es r&rrsj8*---%   I & & %*:*::*F*F INN78<w+O+O+O+W#X#XYYEE W^^E " " INN5))EE .. , C((+ (88 ws||44    BjoBB@BB sy / /  D    L ' ' . .E MM% E LsAE!! F+FFct|ttfrMt|r0t|dttfrfd|DSfd|DSt |S)aLoads images, handling different levels of nesting. Args: images: A single image, a list of images, or a list of lists of images to load. timeout: Timeout for loading images. Returns: A single image, a list of images, a list of lists of images. rc,g|]}fd|DS)c2g|]}t|SrrrOrGrs r&rXz*load_images...s&QQQEZw777QQQr%r$)rO image_grouprs r&rXzload_images..s/lllVaQQQQ[QQQlllr%c2g|]}t|Srrrs r&rXzload_images..s&KKK5Jug666KKKr%r)r2r\rerr)rKrs `r& load_imagesrs&4-((3 v;; L:fQi$?? Llllleklll lKKKKFKKK K&'2222r% do_rescalerescale_factor do_normalize image_mean image_stddo_padpad_sizedo_center_crop crop_size do_resizesizeresamplePILImageResampling interpolationrc |r|td|r|td|r||td|r|td| | td| r| | | tddSdSdS)a Checks validity of typically used arguments in an `ImageProcessor` `preprocess` method. Raises `ValueError` if arguments incompatibility is caught. Many incompatibilities are model-specific. `do_pad` sometimes needs `size_divisor`, sometimes `size_divisibility`, and sometimes `size`. New models and processors added should follow existing arguments when possible. Nz=`rescale_factor` must be specified if `do_rescale` is `True`.zgDepending on the model, `size_divisor` or `pad_size` or `size` must be specified if `do_pad` is `True`.zP`image_mean` and `image_std` must both be specified if `do_normalize` is `True`.z<`crop_size` must be specified if `do_center_crop` is `True`.zbOnly one of `interpolation` and `resample` should be specified, depending on image processor type.zO`size` and `resample/interpolation` must be specified if `do_resize` is `True`.)rD) rrrrrrrrrrrrrs r&validate_preprocess_argumentsrs.Zn,XYYY  (" u   m+y/@klllY)+WXXX X%9 p   l$*0D Hajkkkll0D0DHaHar%ceZdZdZdZddZdZdejde e e fdejfd Z dd Z d ZddZddZdZdZddZdS)ImageFeatureExtractionMixinzD Mixin that contain utilities for preparing image features. ct|tjjtjfs/t |s"t dt|ddSdS)Nz Got type zU which is not supported, only `PIL.Image.Image`, `np.ndarray` and `torch.Tensor` are.)r2r3r4r]r^rrDrEselfrGs r&_ensure_format_supportedz4ImageFeatureExtractionMixin._ensure_format_supported>sq%#)/2:!>?? X]H^H^ &DKK&&&     r%Nc||t|r|}t|tjr|%t|jdtj}|jdkr&|j ddvr| ddd}|r|dz}| tj }tj|S|S)a" Converts `image` to a PIL Image. Optionally rescales it and puts the channel dimension back as the last axis if needed. Args: image (`PIL.Image.Image` or `numpy.ndarray` or `torch.Tensor`): The image to convert to the PIL Image format. rescale (`bool`, *optional*): Whether or not to apply the scaling factor (to make pixel values integers between 0 and 255). Will default to `True` if the image type is a floating type, `False` otherwise. Nrrqrrr)rrr<r2r]r^flatfloatingrur transposeastypermr3r4 fromarray)rrGrescales r& to_pil_imagez(ImageFeatureExtractionMixin.to_pil_imageEs %%e,,, 5 ! ! "KKMME eRZ ( ( .$UZ]BK@@zQ5;q>V#;#;1a00 $ LL**E9&&u-- - r%c||t|tjjs|S|dS)z Converts `PIL.Image.Image` to RGB format. Args: image (`PIL.Image.Image`): The image to convert. r)rr2r3r4rrs r& convert_rgbz'ImageFeatureExtractionMixin.convert_rgbcsE %%e,,,%11 L}}U###r%rGscalerjc6||||zS)z7 Rescale a numpy image by scale amount )r)rrGr s r&r z#ImageFeatureExtractionMixin.rescaleqs" %%e,,,u}r%Tc||t|tjjrt j|}t |r|}|%t|jdtj n|}|r3| | tj d}|r"|j dkr|ddd}|S)a Converts `image` to a numpy array. Optionally rescales it and puts the channel dimension as the first dimension. Args: image (`PIL.Image.Image` or `np.ndarray` or `torch.Tensor`): The image to convert to a NumPy array. rescale (`bool`, *optional*): Whether or not to apply the scaling factor (to make pixel values floats between 0. and 1.). Will default to `True` if the image is a PIL Image or an array/tensor of integers, `False` otherwise. channel_first (`bool`, *optional*, defaults to `True`): Whether or not to permute the dimensions of the image to put the channel dimension first. Nrp?rqrr)rr2r3r4r]rrr<rintegerr rfloat32rur)rrGr  channel_firsts r&rz*ImageFeatureExtractionMixin.to_numpy_arrayxs %%e,,, eSY_ - - $HUOOE 5 ! ! "KKMME;B?*UZ]BJ777PW  FLLbj!9!99EEE  -UZ1__OOAq!,,E r%c||t|tjjr|St |r|d}nt j|d}|S)z Expands 2-dimensional `image` to 3 dimensions. Args: image (`PIL.Image.Image` or `np.ndarray` or `torch.Tensor`): The image to expand. rrY)rr2r3r4r unsqueezer] expand_dimsrs r&rz'ImageFeatureExtractionMixin.expand_dimssq %%e,,, eSY_ - - L 5 ! ! 2OOA&&EEN5q111E r%Fc||t|tjjr||d}n|rt|t jr4||t j d}n7t|r(|| d}t|t jrt|t js,t j ||j }t|t js,t j ||j }nt|rddl}t||js;t|t jr|j|}n|j|}t||js;t|t jr|j|}n|j|}|jdkr-|jddvr||ddddfz |ddddfz S||z |z S)a  Normalizes `image` with `mean` and `std`. Note that this will trigger a conversion of `image` to a NumPy array if it's a PIL Image. Args: image (`PIL.Image.Image` or `np.ndarray` or `torch.Tensor`): The image to normalize. mean (`list[float]` or `np.ndarray` or `torch.Tensor`): The mean (per channel) to use for normalization. std (`list[float]` or `np.ndarray` or `torch.Tensor`): The standard deviation (per channel) to use for normalization. rescale (`bool`, *optional*, defaults to `False`): Whether or not to rescale the image to be between 0 and 1. If a PIL image is provided, scaling will happen automatically. T)r rrNrqr)rr2r3r4rr]r^r rrrfloatrrlr;r` from_numpytensorrur)rrGmeanstdr r;s r& normalizez%ImageFeatureExtractionMixin.normalizes6 %%e,,, eSY_ - - ?''t'<> eRZ ( ( ,dBJ// :x~~,,U[99c2:.. 8hsmm**5;77 U # # , LLLdEL11 .dBJ//.+5+D11DD'5<--Dc5<00 ,c2:..,*%*3//CC&%,s++C :??u{1~77DD$//3qqq$}3EE EDLC' 'r%c||n tj}||t|tjjs||}t|trt|}t|tst|dkr|r*t|tr||fn|d|df}n|j \}}||kr||fn||f\}} t|tr|n|d} || kr|S| t| | z|z } } |8|| krtd|d|| |krt|| z| z |} } ||kr| | fn| | f}| ||S)a Resizes `image`. Enforces conversion of input to PIL.Image. Args: image (`PIL.Image.Image` or `np.ndarray` or `torch.Tensor`): The image to resize. size (`int` or `tuple[int, int]`): The size to use for resizing the image. If `size` is a sequence like (h, w), output size will be matched to this. If `size` is an int and `default_to_square` is `True`, then image will be resized to (size, size). If `size` is an int and `default_to_square` is `False`, then smaller edge of the image will be matched to this number. i.e, if height > width, then image will be rescaled to (size * height / width, size). resample (`int`, *optional*, defaults to `PILImageResampling.BILINEAR`): The filter to user for resampling. default_to_square (`bool`, *optional*, defaults to `True`): How to convert `size` when it is a single int. If set to `True`, the `size` will be converted to a square (`size`,`size`). If set to `False`, will replicate [`torchvision.transforms.Resize`](https://pytorch.org/vision/stable/transforms.html#torchvision.transforms.Resize) with support for resizing only the smallest edge and providing an optional `max_size`. max_size (`int`, *optional*, defaults to `None`): The maximum allowed for the longer edge of the resized image: if the longer edge of the image is greater than `max_size` after being resized according to `size`, then the image is resized again so that the longer edge is equal to `max_size`. As a result, `size` might be overruled, i.e the smaller edge may be shorter than `size`. Only used if `default_to_square` is `False`. Returns: image: A resized `PIL.Image.Image`. Nrrz max_size = zN must be strictly greater than the requested size for the smaller edge size = )r)rBILINEARrr2r3r4r r\rerrrrDresize) rrGrrdefault_to_squaremax_sizerrshortlongrequested_new_short new_shortnew_longs r&r z"ImageFeatureExtractionMixin.resizes< (3889K9T %%e,,,%11 -%%e,,E dD ! ! ;;D dC  [CIINN  ['1$'<'<Td||47DQRGBT % v16&ufoovuo t.8s.C.C&Pdda#/// L&93?RUY?Y\a?a;b;b8 '#666(G(GG@DGG (**.1(Y2F2Q.R.RT\8 05 8,,hPYEZ||D8|444r%cV||t|ts||f}t|st|tjrN|jdkr||}|jddvr|jddn|jdd}n|j d|j df}|d|dz dz}||dz}|d|dz dz}||dz}t|tj j r| ||||fS|jddv}|sWt|tjr| ddd}t|r|ddd}|dkr-||dkr!|dkr||dkr|d||||fS|jddt|d|dt|d|dfz} t|tjrt j|| } n$t|r|| } | d|dz dz} | |dz} | d |dz dz} | |dz}|| d| | | |f<|| z }|| z }|| z }|| z }| dtd|t%| jd|td|t%| jd |f} | S) a Crops `image` to the given size using a center crop. Note that if the image is too small to be cropped to the size given, it will be padded (so the returned result has the size asked). Args: image (`PIL.Image.Image` or `np.ndarray` or `torch.Tensor` of shape (n_channels, height, width) or (height, width, n_channels)): The image to resize. size (`int` or `tuple[int, int]`): The size to which crop the image. Returns: new_image: A center cropped `PIL.Image.Image` or `np.ndarray` or `torch.Tensor` of shape: (n_channels, height, width). rrrrN.r)rr)rr2rerr]r^rurrrr3r4croprpermutero zeros_like new_zerosrn)rrGr image_shapetopbottomleftrightr new_shape new_imagetop_pad bottom_padleft_pad right_pads r& center_cropz'ImageFeatureExtractionMixin.center_crop#sd %%e,,,$&& $1a7{1~- bMKN2q8{1~- AF #wz)8I+==> w'   Qs9?2#6???QPST]TcdfTginPoPoAo o r%c||t|tjjr||}|dddddddfS)a Flips the channel order of `image` from RGB to BGR, or vice versa. Note that this will trigger a conversion of `image` to a NumPy array if it's a PIL Image. Args: image (`PIL.Image.Image` or `np.ndarray` or `torch.Tensor`): The image whose color channels to flip. If `np.ndarray` or `torch.Tensor`, the channel dimension should be first. Nr)rr2r3r4rrs r&flip_channel_orderz.ImageFeatureExtractionMixin.flip_channel_orderns` %%e,,, eSY_ - - /''..ETTrT111aaaZ  r%rc||ntjj}||t |tjjs||}|||||||S)a Returns a rotated copy of `image`. This method returns a copy of `image`, rotated the given number of degrees counter clockwise around its centre. Args: image (`PIL.Image.Image` or `np.ndarray` or `torch.Tensor`): The image to rotate. If `np.ndarray` or `torch.Tensor`, will be converted to `PIL.Image.Image` before rotating. Returns: image: A rotated `PIL.Image.Image`. N)rexpandcenter translate fillcolor)r3r4NEARESTrr2r rotate)rrGanglerr<r=r>r?s r&rAz"ImageFeatureExtractionMixin.rotates} (3889J %%e,,,%11 -%%e,,E|| HVFicl   r%r1)NT)F)NTN)NrNNN)rr r!__doc__rr r r]r^rrrr rrrr r8r:rAr$r%r&rr9s< $ $ $RZeSj0Abj@(2(2(2(2(hA5A5A5A5FIIIV!!!"      r%rannotation_formatsupported_annotation_formatsc||vrtdtd||tjurt |std|tjurt |stddSdS)NzUnsupported annotation format: z must be one of zInvalid COCO detection annotations. Annotations must a dict (single image) or list of dicts (batch of images) with the following keys: `image_id` and `annotations`, with the latter being a list of annotations in the COCO format.zInvalid COCO panoptic annotations. Annotations must a dict (single image) or list of dicts (batch of images) with the following keys: `image_id`, `file_name` and `segments_info`, with the latter being a list of annotations in the COCO format.)rDformatr(r+rr,r)rDrErs r&validate_annotationsrHs  <<<q6qqSoqqrrr,;;;/ << B  ,:::.{;; M ;:  r%valid_processor_keyscaptured_kwargsct|t|}|r5d|}td|ddSdS)Nz, zUnused or unrecognized kwargs: rt)set differencejoinrr)rIrJ unused_keysunused_key_strs r&validate_kwargsrQspo&&11#6J2K2KLLKL;//JJJJKKKKKLLr%T)frozenceZdZUdZdZeeed<dZeeed<dZ eeed<dZ eeed<dZ eeed<dZ eeed<d Z dS) SizeDictz> Hashable dictionary to store image size information. Nrr longest_edge shortest_edgerrcht||rt||Std|d)NzKey z not found in SizeDict.)hasattrgetattrKeyError)rkeys r& __getitem__zSizeDict.__getitem__s> 4   &4%% %:c:::;;;r%)rr r!rCrrr__annotations__rrUrVrrr\r$r%r&rTrTs!FHSM   E8C="&L(3-&&&#'M8C=''' $J $$$#Ix}###<<<<rrs  $$$$$$!!!!!!""""""""                          --!! -<<<<<<  &(9(G  "$5$9  '):)C  &(9(A  &(9(A  &(9(A + ''+-'LLL  H % %rz>48I3JDQSQ[L\^bcq^rr | $$$$$|$$$ 99999\999 c5c4:!5667FFF  ? ? ?wwwEDEEEE,,,    52:5$5555$$$D 4U4;=O8O3P UY     $sE$+ "1l+,1l/01l1l1l1lj\ \ \ \ \ \ \ \ ~ '"'(8#(=">d 2L$s)Ld3iLLLL $<<<<<<<<<