3jeI.SrSSKrSSKrSSKJr SSKJr SSKJ r J r SSK J r J r SSKJrJrJrJrJr SS KJrJr SS KJrJrJrJr \R8"\5rS \\\\\-\-S \\\4S jr "SS\SS9r!\"SS\55r"S/r#g)zImage processor class for Fuyu.N) functional)TorchvisionBackend) BatchFeature get_size_dict)group_images_by_shapereorder_images) ImageInputPILImageResamplingSizeDictis_valid_imagemake_list_of_images) ImagesKwargsUnpack) TensorTypeauto_docstringloggingrequires_backendsimagesreturnc[U5(aU//$[U[5(a[SU55(aU$[U[5(aUVs/sHn[ U5PM sn$[ S5es snf)Nc3B# UHn[U[5v M g7fN) isinstancelist).0images h/home/wildlama/miniconda3/lib/python3.13/site-packages/transformers/models/fuyu/image_processing_fuyu.py .make_list_of_list_of_images..1s'TVE 5$(?(?VszHimages must be a list of list of images or a list of images or an image.)r rrallr ValueError)rrs rmake_list_of_list_of_imagesr#+sufz&$C'TV'T$T$T &$8>?u#E*?? _ ``@sA?c>\rSrSr%Sr\S-\S'\\S'\\S'Sr g)FuyuImagesKwargs:a patch_size (`dict[str, int]`, *optional*, defaults to `{"height": 30, "width": 30}`): Dictionary in the format `{"height": int, "width": int}` specifying the size of the patches. padding_value (`float`, *optional*, defaults to 1.0): The value to pad the image with. padding_mode (`str`, *optional*, defaults to "constant"): The padding mode to use when padding the image. N patch_size padding_value padding_mode) __name__ __module__ __qualname____firstlineno____doc__r __annotations__floatstr__static_attributes__r*rr%r%:s4r4r%F)totalc^\rSrSrSrSSS.rSSS.r\Rr Sr Sr Sr Sr S rS rSrS r/S Qr\rS \\4U4S jjrS5S\S\S\4SjjrS6S\R6S\SSS\S\R64 U4SjjjrS\SS\S\SSS\S\ S\S\ \\ -S-S\ \\ -S-S \S-S!\ S-S"\!S-S#\S-S$\!\"-S-S\#4S%jr$S7S&\S'\S(\S-S\4S)jjr%S7S\R6S(\S-S\R64S*jjr&S7S+\R6S,\R6S-\R6S.\R6S/\S0\S1\S(\'\!\4S-S\#4S2jjr(S7S(\'\!\4\-S-S\'4U4S3jjjr)S4r*U=r+$)8FuyuImageProcessorITi8iheightwidthg?constantg?gp?rimage_input_ids image_patchesimage_patch_indices_per_batch#image_patch_indices_per_subsequencekwargsc &>[TU]"S0UD6 g)Nr*)super__init__)selfrC __class__s rrFFuyuImageProcessor.__init__`s "6"r4rexpected_ndimsrc:URU5n[U5$r) fetch_imagesr#)rGrrJs r_prepare_images_structure,FuyuImageProcessor._prepare_images_structurecs ""6**622r4Nrsizeresamplez7PILImageResampling | tvF.InterpolationMode | int | None antialiasc >Uc[RnURSSupgURURpXy::aXh::aU$X- n X- n [ X5n [ Xl-5n [ X|-5n[TU]!U[XS9X4S9$)av Resize an image to fit within `(size.height, size.width)` while maintaining aspect ratio. Only resizes if the image is larger than the target size. Args: image (`torch.Tensor`): Image to resize. size (`SizeDict`): Dictionary in the format `{"height": int, "width": int}` specifying the max size of the output image. resample (`PILImageResampling | tvF.InterpolationMode | int`, *optional*, defaults to `PILImageResampling.BILINEAR`): Resampling filter to use when resizing the image. antialias (`bool`, *optional*, defaults to `True`): Whether to apply antialiasing when resizing. Nr9)rPrQ) r BILINEARshaper:r;minintrEresizer )rGrrOrPrQrC image_height image_width target_height target_widthheight_scale_factorwidth_scale_factoroptimal_scale_factor new_height new_widthrHs rrXFuyuImageProcessor.resizeks*  )22H$)KK$4! &*kk4::|  &<+HL+:)7"#6K<=  :; w~ 8:?(  r4z torch.Tensor do_resize do_rescalerescale_factor do_normalize image_mean image_stddo_padr(r)disable_groupingreturn_tensorsc UVs/sHnU(dM USRSSPM nn[XSS9unn0nUR5H"unnU(aURUX4S9nUUU'M$ [ UUSS9nUVs/sHnU(dM USRSSPM nnUVs/sH nUS/PM nnUVs/sH nUS/PM nn[ UU5VVs/sHunnUSUS- /PM nnnU (aUR UUU U U SS9n[UU SS9unn0nUR5HunnURUXVXxU 5nUUU'M! [ UUSS9n [R"U V!s/sH"n!U!(dM [R"U!5PM$ sn!5n"[U"UUUS .US /S 9$s snfs snfs snfs snfs snnfs sn!f) NrrST)rj is_nested)rrOrP)rm)pad_size fill_valuer)rjrm)rimage_unpadded_heightsimage_unpadded_widthsimage_scale_factorsoverflowing_values)data tensor_typeskip_tensor_conversion) rUritemsrXr zippadrescale_and_normalizetorchstackr)#rGrrcrOrPrdrerfrgrhrir(r)rjrkrC batch_imageoriginal_image_sizesgrouped_imagesgrouped_images_indexresized_images_groupedrUstacked_imagesresized_images image_sizes image_sizerqrr original_size resized_sizersprocessed_images_groupedprocessed_imagesbatch images_tensors# r _preprocessFuyuImageProcessor._preprocesssQ&NTcVkWb 9 A 4 4RS 9Vc/D 0 ,,"$%3%9%9%; !E>!%>!`,: "5 )&<((>@T`deDRbN[Va0{1~++BC0N bDO!PKj:a=/K!PCN O;Z*Q-; O033G/U /U+ |!_}Q/ / 0/U   !XX()!1 &N0E -=0 ,,$& %3%9%9%; !E>!77 LV_N/= $U + &< **BDXdhi EU$_EUEY^%7U[[%7EU$_` '*@)>':  '$8#9  U dc!P O 6%`s4 F<F< GG+GG "G; G GrYrZr'c<UcA[UR[5(a URnO[S0URD6nURURpTX-S:wa[ SU<SU35eX%-S:wa[ SU<SU35eX-nX%-nXg-nU$)a: Calculate number of patches required to encode an image. Args: image_height (`int`): Height of the image. image_width (`int`): Width of the image. patch_size (`SizeDict`, *optional*): Dictionary in the format `{"height": int, "width": int}` specifying the size of the patches. rz image_height=z must be divisible by z image_width=r*)rr'r r:r;r") rGrYrZr' patch_height patch_widthnum_patches_per_dim_hnum_patches_per_dim_w num_patchess rget_num_patches"FuyuImageProcessor.get_num_patchess  $//844!__ %88 $.$5$5z7G7Gk  &! + .D\NST T  $ ) ~-CK=QR R , < + :+C r4c[US/5 UcA[UR[5(a URnO[S0URD6nURUR pCUR upV nURSX35nURSXD5n U R5n U RXVSX45n U RSSSSS5n U RUSXc-U-5n U $) a? Convert an image into a tensor of patches using PyTorch's unfold operation. Args: image (`torch.Tensor`): Image to convert. Shape: [batch, channels, height, width] patch_size (`SizeDict`, *optional*): Dictionary in the format `{"height": int, "width": int}` specifying the size of the patches. r|rrrnr*) rrr'r r:r;rUunfold contiguousviewpermutereshape) rGrr'rr batch_sizechannels_unfolded_along_heightpatchess rpatchify_image!FuyuImageProcessor.patchify_images $ *  $//844!__ %88 $.$5$5z7G7Gk%*[[" a % Q K'..q+K$$&,,zRS//!Q1a0//*b(2IK2WXr4 image_input image_presentimage_unpadded_himage_unpadded_wimage_placeholder_idimage_newline_idvariable_sizedc `[US/5 UcB[UR[5(a URnO6[S0URD6nO [U[5(d [S0UD6nURUR p/n /n /n [ URS5GHbn/n/n[ URS5GHnX.U4(GaXU4nURSURSnnU(af[U[R"X>U4U - 5U -5n[U[R"XNU4U - 5U -5nUSS2SU2SU24nUUnnURUUUS9n[R"U/U[RURS9nUR!UR#S5US9R%S5nUURS:XdeU(a{UR'S UU -5n[R"URSS/U[RURS9n[R("UU/SS 9nUR'S 5nU R+U/5 UR+U5 UR+U5 GMUR+[R,"/[RURS95 GM U R+U5 U R+U5 GMe /n/nU GHnSn/n/n UHn!U!U:Hn"[R."U"5n[R0"U[R2U!RS9R5U!5n#[R6"U!S 5n$[R6"U!S 5n%[R8"U"S S 9Sn&U#U-U$U&'U#U%U&'UR+U$5 U R+U%5 UU- nM UR+U5 UR+U 5 GM [;U U U UUS .S9$)a Process images for model input. In particular, variable-sized images are handled here. Args: image_input (`torch.Tensor` of shape [batch_size, subsequence_size, num_channels, height, width]): Tensor of images padded to model input size. image_present (`torch.Tensor` of shape [batch_size, subsequence_size, num_images]): Tensor of 1s and 0s indicating whether an image is present. image_unpadded_h (`torch.Tensor` of shape [batch_size, subsequence_size]): Tensor of unpadded image heights. image_unpadded_w (`torch.Tensor` of shape [batch_size, subsequence_size]): Tensor of unpadded image widths. image_placeholder_id (int): The id of the image placeholder token. Comes from an associated tokenizer. image_newline_id (int): The id of the image newline token. Comes from an associated tokenizer. variable_sized (bool): Whether to process images as variable-sized. patch_size (`dict[str, int]`, *optional*): Size of the patches. r|Nrrnr)rYrZr')dtypedevice)rr'r)dimT)as_tupler>)rur*)rrr'r r:r;rangerUrVmathceilrr|fullint32rr unsqueezesqueezercatappendtensor count_nonzeroarangeint64type_as full_likenonzeror)'rGrrrrrrrr'rrrbatch_image_patchesbatch_image_input_ids batch_indexr?r@ subseq_indexrrYrZnew_hnew_wrtensor_of_image_idsr newline_idsrArBsample_image_input_ids index_offsetper_batch_indicesper_subsequence_indicessubseq_image_input_ids patches_maskindicesindices_in_stream_per_batch!indices_in_stream_per_subsequence patches_indss' rpreprocess_with_tokenizer_info1FuyuImageProcessor.preprocess_with_tokenizer_infos1@ $ *  $//844!__ %88 J11!/J/J$.$5$5z7G7Gk+-8::< !2!21!56K OM %k&7&7&: ;  l!:;;'\(ABE05 A A+L%!$( II&6L7P&QT`&`adpp!!$' II&6L7P&QT_&_`cnn!!&a%%&7 8495k "&"6"6%1{Wa#7#K+0**$ ';5;;WbWiWi+'#118JWa1bjjklmG&'--*::::%.A.I.I"k]hNh.i+&+jj066q91=,"'++#.#5#5 ' /4ii9Lk8Z`a.b+.A.I.I".M+MM5'*#**+>?!((1#**5<<%++VaVhVh+ijU!TVX.Y+49OODZ\^4_1$}}\DI!L [TU]"S0UD6nUb([U[5(d[S0[ USS9D6nXS'U$)z1 Process Fuyu-specific kwargs before validation. r') param_namer*)rE_standardize_kwargsrr r)rGr'rCrHs rr&FuyuImageProcessor._standardize_kwargssJ,6v6  !*Z*J*J!WM*$VWJ)| r4r*)r)NTr),r+r,r-r.rcrOr'r rTrPrir(r)rfrgrhrdremodel_input_namesr% valid_kwargsrrFr rWrMr|Tensorr boolrXrr1r2rrrrrdictrrr3 __classcell__)rHs@rr7r7IsIT *D,J!**H FMLLJIJN$L#(8!9#  333  3OS & ||& & L &  &  & & PF ^$F F  F L F  F F F DK'$.F 4;&-F t F t|F DjF +F j(4/F " #F PCcxZ^jm4ELLhoY^YeYeH-1D \\D ||D  ,, D  ,, D " D D D cNT)D  D P8< cNX-4    r4r7)$r/rr|torchvision.transforms.v2rtvFimage_processing_backendsrimage_processing_utilsrrimage_transformsrr image_utilsr r r r rprocessing_utilsrrutilsrrrr get_loggerr+loggerrr#r%r7__all__r*r4rrs& 7;AE5   H % a j! "T*%5 5 B a $z  a |5 X+XXv r4