+ j|}aR7t50t^RIt^RIHt^RIHt^RIHtHtH t H t ^RI t ^RI H t ^RIHtHt^RIHtHt^RIHt^RIHtHtHt^R IHt^R IHtHt.R8Ot!RR] 4t !RR]4t!!RR] PD4t#!RR] PD4t$!RR ] PD4t%RRlt&R]/t']^k/]'CRRRR /Ct(!R!R ]4t)!R"R ]4t*!R#R]4t+!R$R]4t,!R%R]4t-]!4]!R&])P\3R'7R(RR)R*/R+R,ll44t/]!4]!R&]*P\3R'7R(RR)R*/R-R.ll44t0]!4]!R&]+P\3R'7R(RR)R*/R/R0ll44t1]!4]!R&],P\3R'7R(RR)R*/R1R2ll44t2]!4]!R9R'7R(RR)R*/R3R4ll44t3R:R5R6llt4R#);N) OrderedDict)partial)AnyCallable NamedTupleOptional)Conv2dNormActivationMLP)ImageClassificationInterpolationMode)_log_api_usage_once)register_modelWeights WeightsEnum)_IMAGENET_CATEGORIES)_ovewrite_named_paramhandle_legacy_interfaceVisionTransformerViT_B_16_WeightsViT_B_32_WeightsViT_L_16_WeightsViT_L_32_WeightsViT_H_14_Weightsc\a]tRt^ to]P t]PtV3Rlt Rt Vt R#)ConvStemConfigc<V^8dQh/S[;R&S[;R&S[;R&S[RS[P3,;R&S[RS[P3,;R&#) out_channels kernel_sizestride. norm_layeractivation_layer)intrnnModule)format __classdict__s"f/home/luca/screen-checking/.venv/lib/python3.14/site-packages/torchvision/models/vision_transformer.py __annotate__ConvStemConfig.__annotate__ s[ Kbii(9  sBII~.8 N) __name__ __module__ __qualname____firstlineno__r$ BatchNorm2dr!ReLUr"__annotate_func____static_attributes____classdictcell__)r's@r(rr s!,.>>J13 r+rcRaa]tRt^(toRt^tV3RlV3RlltV3RltRtVt V;t #)MLPBlockzTransformer MLP block.c,<V^8dQhRS[RS[RS[/#)rin_dimmlp_dimdropout)r#float)r&r's"r(r)MLPBlock.__annotate__-s"66s6S656r+c<\SV`WV.\PRVR7VP 4Fp\ V\P 4'gK%\PPVP4VPfK^\PPVPRR7K R#)N)r"inplacer;ư>std) super__init__r$GELUmodules isinstanceLinearinitxavier_uniform_weightbiasnormal_)selfr9r:r;m __class__s&&&& r(rDMLPBlock.__init__-s} 6!2RWWVZdklA!RYY''''166%GGOOAFFO5  r+c <VPRR4pVeV^8dX\^4FHp RF?p V RV ^, RV 2p V ^V , RV 2p W9gK,VPV 4W&KA KJ \S V`VVVVVVV4R#)versionNlinear_.)rKrL)getrangepoprC_load_from_state_dict)rN state_dictprefixlocal_metadatastrict missing_keysunexpected_keys error_msgsrSitypeold_keynew_keyrPs&&&&&&&& r(rYMLPBlock._load_from_state_dict6s!$$Y5 ?gk1X.D!'!uAdV\PJPWVP0PXPT4\]VPDR 4'd\GVPDP^\P@4'dVPDP^P`p\PJPMVPDP^PN\PPR!^V, 4R 7\PJPWVPDP^PT4\GVPDPb\P@4'd}\PJPWVPDPbPN4\PJPWVPDPbPT4R#R#)rz&Input shape indivisible by patch size!N conv_bn_relu_) in_channelsrrr r!r" conv_last)rrr)rrrr head pre_logitsactrArg@)meanrB)2rCrDr rsrrrrpr:rqr;rrr!r$r enumerate add_moduler rrr r"Conv2d conv_projrzeros class_tokenrencoderrrrHTanhheadsrGrrI trunc_normal_rKmathsqrtrLzeros_rrMhasattrr in_featuresr)rNrrrrorpr:r;rqrrr!rseq_proj prev_channelsraconv_stem_layer_configr heads_layersfan_inrPs&&&&&&&&&&&&& r(rDVisionTransformer.__init__s? D! j-24\]$$$ !2 &#6 $  (}}HM-67H-I))###A3'($1%;%H%H$:$F$F5<<#9#D#D)?)P)P  !7 C C .J   RYY=ghi )1NYYJWaDN!.14 << Aq*(EFa            %4?M  &#%99Z#EL )+:)SL &"$'')L #%99-@#NL ]]<0 dnnbii 0 0^^//$..2L2LQ2OORVR`R`RlRlmnRooF GG ! !$.."7"7TYYq6z=R ! S~~"".t~~223 ^^ % % 1jAYAY[][d[d6e6e GGOO((//ctyyt~~OgOgOtOtIt?u  ~~'',,8t~~77<<= 4::| , ,DJJ">DIIaRXjDY ! Z GGNN4::0055 6 djjooryy 1 1 GGNN4::??11 2 GGNN4::??// 0 2r+cN<V^8dQhRS[PRS[P/#)rrreturnr)r&r's"r(r)r s# r+cVPwr#rEVPp\P!W@P8HRVP RV R24\P!WPP8HRVP RV R24WF,pWV,pVP V4pVP W PWx,4pVP^^^4pV#)zWrong image height! Expected z but got !zWrong image width! Expected ) rrrsrrrreshaperppermute) rNrnchwpn_hn_ws && r(_process_input VisionTransformer._process_input sWW a OO a??*.KDOOK\\efgehhi,jk a??*.J4??J[[defdggh,ijff NN1  IIa#) 4 IIaA r+c4<V^8dQhRS[P/#)rrr)r&r's"r(r)r!sr+cVPV4pVP^,pVPPVRR4p\P !W1.^R7pVP V4pVR,pVPV4pV#)rr)NNNr)rrrexpandrscatrr)rNrrbatch_class_tokens&& r(rVisionTransformer.forward!sz    " GGAJ!,,33Ar2> II(,! 4 LLO dG JJqMr+)rqrrr;rrrprr:r!rrrr)r-r.r/r0rfrr$rrDrrr4r5rhris@@r(rrsME#&-15`_'s training recipe. url transformsrz>https://download.pytorch.org/models/vit_b_16_swag-9ac1b537.pthr resize_size interpolationi^-g~jtSU@giX@gˡEK@g|?5^t@ These weights are learnt via transfer learning by end-to-end fine-tuning the original `SWAG `_ weights on ImageNet-1K data. zAhttps://download.pytorch.org/models/vit_b_16_lc_swag-4e70ced5.pth+https://github.com/pytorch/vision/pull/5793gbX9xT@gQ X@ These weights are composed of the original frozen `SWAG `_ trunk weights and a linear classifier learnt on top of them trained on ImageNet-1K data. r,Nr )rrr-r.r/r0rrr _COMMON_META IMAGENET1K_V1r BICUBIC_COMMON_SWAG_METAIMAGENET1K_SWAG_E2E_V1IMAGENET1K_SWAG_LINEAR_V1DEFAULTr4r,r+r(rr_s_ G.#>  (  f VV  F '  M,% L +33    (  VV  F '  4!( O +33    C (  VV  F '  !6Gr+cp]tRtRt]!R]!]^R7/]CRRRRRRR R R R R R//RRRRRR/CR7t]t Rt R#)riz9https://download.pytorch.org/models/vit_b_32-d86f8d99.pthrri1BrrzNhttps://github.com/pytorch/vision/tree/main/references/classification#vit_b_32rrrg|?5^R@rgW@rgA`Т@rgl u@rrrr,Nr  r-r.r/r0rrr r rrr4r,r+r(rrsq G.#>  (  f VV  E '  M,Gr+cJ]tRtRt]!R]!]^^R7/]CRRRR&RRR R R R R R//RRRRRR/CR7t]!R]!]RR] PR7/] CRRRR'R R R RR R//RRRRRR/CR7t ]!R]!]^^] PR7/] CRR RRRR&R R R R!R R"//RRRRRR#/CR7t ]tR$tR%#)(riz9https://download.pytorch.org/models/vit_l_16-852ce7e3.pth)rrri#rrzNhttps://github.com/pytorch/vision/tree/main/references/classification#vit_l_16rrrg|?5^S@rgFԨW@rgףp= N@rg;O$@ra These weights were trained from scratch by using a modified version of TorchVision's `new training recipe `_. rz>https://download.pytorch.org/models/vit_l_16_swag-4f3808c9.pthri0gjtV@gT㥛ĠX@gƟv@gy&11@rzAhttps://download.pytorch.org/models/vit_l_16_lc_swag-4d563306.pthrgMbXIU@g^I [X@r r,Nr )rrr r,r+r(rrsa G.#3O  )  f VV  F (  M.% L +33    )  VV  G (  4!( O +33    C )  VV  F (  !6Gr+cp]tRtRt]!R]!]^R7/]CRRRRRRR R R R R R//RRRRRR/CR7t]t Rt R#)riz9https://download.pytorch.org/models/vit_l_32-c7638314.pthrri[ErrzNhttps://github.com/pytorch/vision/tree/main/references/classification#vit_l_32rrrg|?5>S@rgGzDW@rgK7.@rgE@rrrr,Nr rr,r+r(rrsq G.#>  )  f VU  F (  M,Gr+c]tRtRt]!R]!]RR]PR7/] CRRRR RR R R R R //RRRRRR/CR7t ]!R]!]^^]PR7/] CRRRRRR!RR R RR R//RRRRRR/CR7t ] t Rt R#)"ri2z>https://download.pytorch.org/models/vit_h_14_swag-80465313.pthrri%rrrrgS#V@rg#~jX@rg~jŏ@rgK7I@rrrzAhttps://download.pytorch.org/models/vit_h_14_lc_swag-c1eb923e.pthrri@%gZd;OmU@gQnX@g= ףpd@gI k֢@r r,N)rrr )r-r.r/r0rrr r rrrrrr4r,r+r(rr2s$ L +33    )  VV  H (  4!( O +33    C )  VV  G (  !6%Gr+ pretrained)rrrTc^V^8dQhR\\,R\R\R\/#rrrrr)rrrrr)r&s"r(r)r)m0"23d]`evr+c ^\PV4p\R R^R^ R^ RRRRRVR V/VB#) au Constructs a vit_b_16 architecture from `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale `_. Args: weights (:class:`~torchvision.models.ViT_B_16_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.ViT_B_16_Weights` below for more details and possible values. By default, no pre-trained weights are used. progress (bool, optional): If True, displays a progress bar of the download to stderr. Default is True. **kwargs: parameters passed to the ``torchvision.models.vision_transformer.VisionTransformer`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.ViT_B_16_Weights :members: rrrorpr: rrr,)rverifyrrrrs$$,r(vit_b_16r$km(%%g.G              r+c^V^8dQhR\\,R\R\R\/#r)rrrrr)r&s"r(r)r)rr+c ^\PV4p\R R^ R^ R^ RRRRRVR V/VB#) au Constructs a vit_b_32 architecture from `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale `_. Args: weights (:class:`~torchvision.models.ViT_B_32_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.ViT_B_32_Weights` below for more details and possible values. By default, no pre-trained weights are used. progress (bool, optional): If True, displays a progress bar of the download to stderr. Default is True. **kwargs: parameters passed to the ``torchvision.models.vision_transformer.VisionTransformer`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.ViT_B_32_Weights :members: rrrorpr r:r!rrr,)rr"rr#s$$,r(vit_b_32r(r%r+c^V^8dQhR\\,R\R\R\/#r)rrrrr)r&s"r(r)r)rr+c ^\PV4p\R R^R^R^RRRRRVR V/VB#) au Constructs a vit_l_16 architecture from `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale `_. Args: weights (:class:`~torchvision.models.ViT_L_16_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.ViT_L_16_Weights` below for more details and possible values. By default, no pre-trained weights are used. progress (bool, optional): If True, displays a progress bar of the download to stderr. Default is True. **kwargs: parameters passed to the ``torchvision.models.vision_transformer.VisionTransformer`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.ViT_L_16_Weights :members: rrrorpr:rrr,)rr"rr#s$$,r(vit_l_16r-m(%%g.G              r+c^V^8dQhR\\,R\R\R\/#r)rrrrr)r&s"r(r)r)rr+c ^\PV4p\R R^ R^R^RRRRRVR V/VB#) au Constructs a vit_l_32 architecture from `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale `_. Args: weights (:class:`~torchvision.models.ViT_L_32_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.ViT_L_32_Weights` below for more details and possible values. By default, no pre-trained weights are used. progress (bool, optional): If True, displays a progress bar of the download to stderr. Default is True. **kwargs: parameters passed to the ``torchvision.models.vision_transformer.VisionTransformer`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.ViT_L_32_Weights :members: rrrorpr+r:r,rrr,)rr"rr#s$$,r(vit_l_32r1r.r+c^V^8dQhR\\,R\R\R\/#r)rrrrr)r&s"r(r)r)rr+c ^\PV4p\R R^R^ R^RRRRRVR V/VB#) au Constructs a vit_h_14 architecture from `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale `_. Args: weights (:class:`~torchvision.models.ViT_H_14_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.ViT_H_14_Weights` below for more details and possible values. By default, no pre-trained weights are used. progress (bool, optional): If True, displays a progress bar of the download to stderr. Default is True. **kwargs: parameters passed to the ``torchvision.models.vision_transformer.VisionTransformer`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.ViT_H_14_Weights :members: rrrorpir:irrr,)rr"rr#s$$,r(vit_h_14r4r.r+c PV^8dQhR\R\RRR\R\RR/#)rrr model_statezOrderedDict[str, torch.Tensor]interpolation_mode reset_headsr)r#strr)r&s"r(r)r)sLKKKK2K K  K & Kr+c:VR,pVPwrgpV^8wd\RVP 24hW,^,^,p W8wEdGV^,pV ^,p VR ,p VR ,p V P^^^4p \\P !V44p W,V8wd\RW, RV 24hV P ^WV 4p W,p \PPV V VRR7pVP ^W4pVP^^^4p\P!W.^R7pWR&V'dE\4pVP4F$wppVPR4'dKVVV&K& TpV#) awThis function helps interpolate positional embeddings during checkpoint loading, especially when you want to apply a pre-trained model on images with different resolution. Args: image_size (int): Image size of the new model. patch_size (int): Patch size of the new model. model_state (OrderedDict[str, torch.Tensor]): State dict of the pre-trained model. interpolation_mode (str): The algorithm used for upsampling. Default: bicubic. reset_heads (bool): If true, not copying the state of heads. Default: False. Returns: OrderedDict[str, torch.Tensor]: A state dict which can be loaded into the new model. zencoder.pos_embeddingz%Unexpected position embedding shape: zPseq_length is not a perfect square! Instead got seq_length_1d * seq_length_1d = z and seq_length = T)sizemode align_cornersrr)r:NrNr)r:rNNr)r ValueErrorrr#rrrr$ functional interpolatersrritems startswith)rrr6r7r8rrrrpnew_seq_lengthpos_embedding_tokenpos_embedding_img seq_length_1dnew_seq_length_1dnew_pos_embedding_imgnew_pos_embeddingmodel_state_copykvs&&&&& r(interpolate_embeddingsrMs* 78M - 3 3A:Av@ATAT@UVWW .14q8N #a !+H5)(3.55aA>DIIj12  (J 6bcpdAcCCUV`Uab  .55aTab&4!# 9 9 "# !:! !6 = =a \!6 = =aA F!II':&RXYZ/@+, AL #))+1||G,,*+$Q',+K r+cbV^8dQh/^\9d\\\3,;R&#)rr )__conditional_annotations__dictr9r)r&s"r(r)r)s& f d38ng  r+) rrrrrrr$r(r-r1r4)rN)bicubicF)6rOr collectionsr functoolsrtypingrrrrrstorch.nnr$ops.miscr r transforms._presetsr r utilsr _apirrr_metar_utilsrr__all__rr7r%rkrrrr rrrrrrrr$r(r-r1r4rMr))rOs@r(r]sE #66 0H'66'B 9Z9+ s+ \!299!H#9bii#9LQ QhB&   8 KL{L^{4M{M`{46%{6%r,0@0N0N!OPdTQ@,0@0N0N!OPdTQ@,0@0N0N!OPdTQ@,0@0N0N!OPdTQ@!56dT7@KKr+