sg?U dZddlmZmZmZddlZddlZddlmZddlm Z m Z m Z ddl m Z ddlmZmZmZmZdd lmZdd lmZmZmZmZmZdd lmZd d lmZej>e Z!dZ"dZ#gdZ$dZ%dZ&d0dejNde(de)dejNfdZ*GddejVZ,GddejVZ-GddejVZ.GddejVZ/Gdd ejVZ0Gd!d"ejVZ1Gd#d$eZ2d%Z3d&Z4ed'e3Gd(d)e2Z5ed*e3Gd+d,e2Z6ed-e3Gd.d/e2eZ7y)1zPyTorch ConvNext model.)OptionalTupleUnionN)nn)BCEWithLogitsLossCrossEntropyLossMSELoss)ACT2FN)BackboneOutputBaseModelOutputWithNoAttention(BaseModelOutputWithPoolingAndNoAttention$ImageClassifierOutputWithNoAttention)PreTrainedModel)add_code_sample_docstringsadd_start_docstrings%add_start_docstrings_to_model_forwardloggingreplace_return_docstrings) BackboneMixin)ConvNextConfigrzfacebook/convnext-tiny-224)rirztabby, tabby catinput drop_probtrainingreturnc|dk(s|s|Sd|z }|jdfd|jdz zz}|tj||j|j z}|j |j||z}|S)aF Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). Comment by Ross Wightman: This is the same as the DropConnect impl I created for EfficientNet, etc networks, however, the original name is misleading as 'Drop Connect' is a different form of dropout in a separate paper... See discussion: https://github.com/tensorflow/tpu/issues/494#issuecomment-532968956 ... I've opted for changing the layer and argument names to 'drop path' rather than mix DropConnect as a layer name and use 'survival rate' as the argument. rr)r)dtypedevice)shapendimtorchrandr r!floor_div)rrr keep_probr" random_tensoroutputs a/var/www/html/venv/lib/python3.12/site-packages/transformers/models/convnext/modeling_convnext.py drop_pathr,:sCx I I [[^  Q 7 7E 5 ELL YYM YYy !M 1F McxeZdZdZd deeddffd ZdejdejfdZ de fdZ xZ S) ConvNextDropPathzXDrop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).Nrrc0t|||_yN)super__init__r)selfr __class__s r+r3zConvNextDropPath.__init__Rs "r- hidden_statescDt||j|jSr1)r,rrr4r6s r+forwardzConvNextDropPath.forwardVs FFr-c8dj|jS)Nzp={})formatr)r4s r+ extra_reprzConvNextDropPath.extra_reprYs}}T^^,,r-r1) __name__ __module__ __qualname____doc__rfloatr3r$Tensorr9strr< __classcell__r5s@r+r/r/OsGb#(5/#T#GU\\GellG-C-r-r/c\eZdZdZdfd Zdej dej fdZxZS)ConvNextLayerNormaALayerNorm that supports two data formats: channels_last (default) or channels_first. The ordering of the dimensions in the inputs. channels_last corresponds to inputs with shape (batch_size, height, width, channels) while channels_first corresponds to inputs with shape (batch_size, channels, height, width). cNt|tjt j ||_tjt j||_||_ ||_ |jdvrtd|j|f|_ y)N) channels_lastchannels_firstzUnsupported data format: ) r2r3r Parameterr$onesweightzerosbiaseps data_formatNotImplementedErrornormalized_shape)r4rSrPrQr5s r+r3zConvNextLayerNorm.__init__cs ll5::.>#?@ LL-=!>? &   #F F%(A$BRBRAS&TU U!1 3r-xrcd|jdk(rWtjjj ||j |j |j|j}|S|jdk(r|j}|j}|jdd}||z jdjdd}||z tj||jzz }|j|}|j ddddf|z|jddddfz}|S)NrIrJrT)keepdim)r )rQr$r functional layer_normrSrMrOrPr rAmeanpowsqrtto)r4rT input_dtypeuss r+r9zConvNextLayerNorm.forwardms     .##..q$2G2GVZV_V_aeaiaijA  !1 1''K Aq$'AQ A##At#4AQ%**Q\22A;'A AtTM*Q.1dD=1IIAr-)ư>rI) r=r>r?r@r3r$rBr9rDrEs@r+rGrG]s( 4  %,, r-rGcZeZdZdZfdZdej dejfdZxZ S)ConvNextEmbeddingszThis class is comparable to (and inspired by) the SwinEmbeddings class found in src/transformers/models/swin/modeling_swin.py. ct|tj|j|j d|j |j |_t|j ddd|_ |j|_y)Nr kernel_sizestriderarJrPrQ) r2r3rConv2d num_channels hidden_sizes patch_sizepatch_embeddingsrG layernormr4configr5s r+r3zConvNextEmbeddings.__init__sr  "   !4!4Q!7VEVEV_e_p_p! +6+>+>q+AtYij"//r- pixel_valuesrc|jd}||jk7r td|j|}|j |}|S)NrzeMake sure that the channel dimension of the pixel values match with the one set in the configuration.)r"rj ValueErrorrmrn)r4rqrj embeddingss r+r9zConvNextEmbeddings.forwardsV#))!, 4,, ,w **<8 ^^J/ r- r=r>r?r@r3r$ FloatTensorrBr9rDrEs@r+rcrc{s*0E$5$5%,,r-rcc\eZdZdZdfd Zdej dejfdZxZ S) ConvNextLayera3This corresponds to the `Block` class in the original implementation. There are two equivalent implementations: [DwConv, LayerNorm (channels_first), Conv, GELU,1x1 Conv]; all in (N, C, H, W) (2) [DwConv, Permute to (N, H, W, C), LayerNorm (channels_last), Linear, GELU, Linear]; Permute back The authors used (2) as they find it slightly faster in PyTorch. Args: config ([`ConvNextConfig`]): Model configuration class. dim (`int`): Number of input channels. drop_path (`float`): Stochastic depth rate. Default: 0.0. c$t|tj||dd||_t |d|_tj|d|z|_t|j|_ tjd|z||_ |jdkDr7tj|jtj |zd nd|_|d kDrt%||_ytj&|_y) Nrr )rfpaddinggroupsrarPrT) requires_gradr)r2r3rridwconvrGrnLinearpwconv1r hidden_actactpwconv2layer_scale_init_valuerKr$rLlayer_scale_parameterr/Identityr,)r4rpdimr,r5s r+r3zConvNextLayer.__init__s iiSa3O *3D9yya#g. &++,yyS#. ,,q0 LL66S9JJZ^ _ " 9BC))4R[[]r-r6rcb|}|j|}|jdddd}|j|}|j|}|j |}|j |}|j |j |z}|jdddd}||j|z}|S)NrrWr r)rpermuternrrrrr,)r4r6rrTs r+r9zConvNextLayer.forwards KK & IIaAq ! NN1  LLO HHQK LLO  % % 1**Q.A IIaAq ! DNN1% %r-)rrurEs@r+rxrxs+  [ U%6%6 5<< r-rxc\eZdZdZdfd Zdej dejfdZxZ S) ConvNextStageaConvNeXT stage, consisting of an optional downsampling layer + multiple residual blocks. Args: config ([`ConvNextConfig`]): Model configuration class. in_channels (`int`): Number of input channels. out_channels (`int`): Number of output channels. depth (`int`): Number of residual blocks. drop_path_rates(`List[float]`): Stochastic depth rates for each layer. c ~t |||k7s|dkDr?tjt |ddtj |||||_ntj|_|xsdg|z}tjt|Dcgc]}t||||c}|_ ycc}w)NrrarJrhrer)rr,) r2r3r SequentialrGridownsampling_layerrrangerxlayers) r4rp in_channels out_channelsrfrgdepthdrop_path_ratesjr5s r+r3zConvNextStage.__init__s  , &&1*&(mm!+4EUV +|U[\'D # ')kkmD #):cUU]mm]bch]i jXYmF PQ@RS j  jsB:r6rcJ|j|}|j|}|Sr1)rrr8s r+r9zConvNextStage.forwards&// >  M2 r-)rWrWrWNrurEs@r+rrs*  U%6%65<<r-rc feZdZfdZ ddej deedeedee e ffdZ xZ S)ConvNextEncoderc (t|tj|_t j d|jt|jj|jDcgc]}|j}}|jd}t|jD]V}|j|}t||||dkDrdnd|j|||}|jj!||}Xycc}w)NrrWr)rrrgrr)r2r3r ModuleListstagesr$linspacedrop_path_ratesumdepthssplittolistrkr num_stagesrappend) r4rprTrprev_chsiout_chsstager5s r+r3zConvNextEncoder.__init__s mmo %q&2G2GV]]I[ \ b bcicpcp q AHHJ  &&q)v(() A))!,G!$$EqqmmA& / 2 E KK  u %H   s8Dr6output_hidden_states return_dictrc|rdnd}t|jD]\}}|r||fz}||}|r||fz}|std||fDSt||S)Nc3&K|] }|| ywr1r).0vs r+ z*ConvNextEncoder.forward..sXq!-Xs)last_hidden_stater6) enumeratertupler )r4r6rrall_hidden_statesr layer_modules r+r9zConvNextEncoder.forwards #7BD(5 8OA|#$58H$H!(7M  8 1]4D D X]4E$FXX X-++  r-)FT) r=r>r?r3r$rvrboolrrr r9rDrEs@r+rrsT.05&*  (( 'tn d^  u44 5  r-rc(eZdZdZeZdZdZdgZdZ y)ConvNextPreTrainedModelz An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained models. convnextrqrxct|tjtjfrm|jj j d|jj|j%|jj jyyt|tjrJ|jj j|jj jdyy)zInitialize the weightsr)rZstdNg?) isinstancerrrirMdatanormal_rpinitializer_rangerOzero_ LayerNormfill_)r4modules r+ _init_weightsz%ConvNextPreTrainedModel._init_weightss fryy"))4 5 MM   & &CT[[5R5R & S{{&   &&('  - KK   " " $ MM   $ $S ).r-N) r=r>r?r@r config_classbase_model_prefixmain_input_name_no_split_modulesrrr-r+rrs' "L"$O() *r-raJ This model is a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior. Parameters: config ([`ConvNextConfig`]): Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights. aF Args: pixel_values (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`): Pixel values. Pixel values can be obtained using [`AutoImageProcessor`]. See [`ConvNextImageProcessor.__call__`] for details. output_hidden_states (`bool`, *optional*): Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for more detail. return_dict (`bool`, *optional*): Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple. zQThe bare ConvNext model outputting raw features without any specific head on top.c eZdZfdZeeeeee de  d de jde ede edeeeffdZxZS) ConvNextModelct||||_t||_t ||_tj|jd|j|_ |jy)Nr|) r2r3rprcrtrencoderrrrklayer_norm_epsrn post_initros r+r3zConvNextModel.__init__Ds`   ,V4&v. f&9&9"&=6CXCXY r-vision) checkpoint output_typermodalityexpected_outputrqrrrcd||n|jj}||n|jj}| td|j |}|j |||}|d}|j |jddg}|s ||f|ddzSt|||jS)Nz You have to specify pixel_valuesrrrrr)r pooler_outputr6) rpruse_return_dictrsrtrrnrZrr6)r4rqrrembedding_outputencoder_outputsr pooled_outputs r+r9zConvNextModel.forwardQs%9$D $++JjJj &1%r?r3rCONVNEXT_INPUTS_DOCSTRINGr_CHECKPOINT_FOR_DOCr_CONFIG_FOR_DOC_EXPECTED_OUTPUT_SHAPEr$rvrrrrr9rDrEs@r+rr?s ++DE&<$. +//3&* " ''" 'tn" d^ " u>> ? " F" r-rz ConvNext Model with an image classification head on top (a linear layer on top of the pooled features), e.g. for ImageNet. ceZdZfdZeeeeee e  d de jde e jde ede edeeeff dZxZS) ConvNextForImageClassificationc0t|||j|_t||_|jdkDr-t j |jd|jnt j|_ |jy)Nrr) r2r3 num_labelsrrrrrkr classifierrros r+r3z'ConvNextForImageClassification.__init__sy   ++%f- FLEVEVYZEZBIIf))"-v/@/@ A`b`k`k`m  r-)rrrrrqlabelsrrrc||n|jj}|j|||}|r |jn|d}|j |}d}||jj |j dk(rd|j_nl|j dkDrL|jtjk(s|jtjk(rd|j_nd|j_|jj dk(rIt} |j dk(r&| |j|j}n| ||}n|jj dk(r=t} | |jd|j |jd}n,|jj dk(rt} | ||}|s|f|ddz} ||f| zS| St!|||j" S) a labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*): Labels for computing the image classification/regression loss. Indices should be in `[0, ..., config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If `config.num_labels > 1` a classification loss is computed (Cross-Entropy). Nrr regressionsingle_label_classificationmulti_label_classificationrrW)losslogitsr6)rprrrr problem_typerr r$longintr squeezerviewrrr6) r4rqrrroutputsrrrloss_fctr*s r+r9z&ConvNextForImageClassification.forwards(&1%r?r3rrr_IMAGE_CLASS_CHECKPOINTrr_IMAGE_CLASS_EXPECTED_OUTPUTr$rvr LongTensorrrrr9rDrEs@r+rr~s ++DE*8$4 +/-1/3&* 3 ''3 ))*3 'tn 3 d^ 3 u:: ; 3 F3 r-rzQ ConvNeXt backbone, to be used with frameworks like DETR and MaskFormer. c eZdZfdZeeeee dde jde e de e defdZ xZS) ConvNextBackbonect||t| |t||_t ||_|jdg|jz|_i}t|j|jD]\}}t|d||<tj||_|j!y)NrrJ)rQ)r2r3_init_backbonercrtrrrk num_featureszip _out_featureschannelsrGr ModuleDicthidden_states_normsr)r4rprrrjr5s r+r3zConvNextBackbone.__init__s   v&,V4&v. #0034v7J7JJ!#&t'9'94==#I g E<):>> from transformers import AutoImageProcessor, AutoBackbone >>> import torch >>> from PIL import Image >>> import requests >>> url = "http://images.cocodataset.org/val2017/000000039769.jpg" >>> image = Image.open(requests.get(url, stream=True).raw) >>> processor = AutoImageProcessor.from_pretrained("facebook/convnext-tiny-224") >>> model = AutoBackbone.from_pretrained("facebook/convnext-tiny-224") >>> inputs = processor(image, return_tensors="pt") >>> outputs = model(**inputs) ```NTrrr) feature_mapsr6 attentions) rprrrtrr6r stage_names out_featuresrr ) r4rqrrrrr6rr hidden_stater*s r+r9zConvNextBackbone.forwards8&1%t77>|L /  0 "_F#=**M%+?-  EI  r-)NN)r=r>r?r3rrrr rr$rBrrr9rDrEs@r+rrsm"++DE>X04&* 9 ll9 'tn9 d^ 9  9 YF9 r-r)rF)8r@typingrrrr$torch.utils.checkpointrtorch.nnrrr activationsr modeling_outputsr r rrmodeling_utilsrutilsrrrrrutils.backbone_utilsrconfiguration_convnextr get_loggerr=loggerrrrrrrBrArr,Moduler/rGrcrxrrrCONVNEXT_START_DOCSTRINGrrrrrr-r+rs)) AA! .22   H %#3'71U\\eTV[VbVb* -ryy - <0)BII)XBII@, bii, ^*o*0  W8 +8 8 v I %<I I X  M . M  M r-