`iz)dZddlmZddlmZddlZddlmZddlmZddl m Z m Z dd l m Z d d lmZee d Gdde Ze GddeZGddejZGddejZGddejZGddejZGddejZe d GddeZddgZdS)zPyTorch ViTMatte model.) dataclass)OptionalN)nn)PreTrainedModel) ModelOutputauto_docstring) load_backbone)VitMatteConfigz4 Class for outputs of image matting models. ) custom_introceZdZUdZdZeejed<dZ eejed<dZ ee ejed<dZ ee ejed<dS)ImageMattingOutputa loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided): Loss. alphas (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`): Estimated alpha values. hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`): Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, + one for the output of each stage) of shape `(batch_size, sequence_length, hidden_size)`. Hidden-states (also called feature maps) of the model at the output of each stage. Nlossalphas hidden_states attentions) __name__ __module__ __qualname____doc__rrtorch FloatTensor__annotations__rrtupler/home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/site-packages/transformers/models/vitmatte/modeling_vitmatte.pyrrs  )-D(5$ %,,,*.FHU& '...8 __class__s rrAzVitMatteBasicConv3x3.__init__Hso I#%     .6;PQQQGII rc||}||}||}|SN)rBrDrFr0 hidden_states rforwardzVitMatteBasicConv3x3.forwardUs;yy.. |44 yy.. r)r8r rrrrrArL __classcell__rGs@rr7r7CsV      rr7c(eZdZdZfdZdZxZS)VitMatteConvStreamzc Simple ConvStream containing a series of basic conv3x3 layers to extract detail features. ctd}|j |jj}|j}t j|_|g|z|_tt|jdz D]H}|j|}|j|dz}|j t|||IdS)Nr ) r@rAbackbone_config num_channelsconvstream_hidden_sizesr ModuleListconvs conv_chansrangelenappendr7)r0r!r:r;iin_chan_ out_chan_rGs rrAzVitMatteConvStream.__init__bs   ! - 0=K5 ]__ &-,6s4?++a/00 Q QAq)HA.I J  268YOO P P P P Q Qrcd|i}|}tt|jD]2}|j||}dt|dzz}|||<3|S)Ndetailed_feature_map_0detailed_feature_map_r )rZr[rXstr)r0r"out_dict embeddingsr]name_s rrLzVitMatteConvStream.forwardusi,l;! s4:'' ) )A&Az22J+c!a%jj8E(HUOOrrMrOs@rrQrQ]sVQQQQQ&rrQc(eZdZdZfdZdZxZS)VitMatteFusionBlockz\ Simple fusion block to fuse features from ConvStream and Plain Vision Transformer. cztt|||dd|_dS)Nr )r=r>)r@rAr7rB)r0r!r:r;rGs rrAzVitMatteFusionBlock.__init__s9 (lST^_``` rctj|ddd}tj||gd}||}|S)Nr8bilinearF) scale_factormode align_cornersr )dim)r functional interpolatercatrB)r0featuresdetailed_feature_mapupscaled_featuresouts rrLzVitMatteFusionBlock.forwardsRM55hQU_ot5uui-/@AqIIIiinn rrMrOs@rrhrhsVaaaaarrhc(eZdZdZfdZdZxZS) VitMatteHeadzJ Simple Matting Head, containing only conv3x3 and conv1x1 layers. c Bt|jd}d}tjtj||dddtj|tjdtj|dddd|_dS)Nrr )r<r=r>Tr) r@rAfusion_hidden_sizesr Sequentialr(r)rE matting_convs)r0r!r: mid_channelsrGs rrAzVitMatteHead.__init__s 04  ] Ik  v) * *c&2P.Q.QTU.U U Uq  ,V44/4]__ & 23f6PPs4/001455  A   % %#! $ 4Q 7$/APQE(:S S!%!5a!e!<    )00rcT||}tt|jD]I}dt t|j|z dz z}|j||||}Jt j||}|S)Nrbr )rrZr[rrcrsigmoidr)r0rsr"detail_featuresr]detailed_feature_map_namers rrLz#VitMatteDetailCaptureModule.forwards//,77s4-..// c cA(?#c$J\F]F]`aFadeFeBfBf(f %,t)!,XG`7abbHHt00::;; rrMrOs@rrrsQ111112rrzX ViTMatte framework leveraging any vision backbone e.g. for ADE20k, CityScapes. c eZdZfdZe d deejdeedeedeejdeef dZ xZ S) VitMatteForImageMattingct|||_t||_t ||_|dSrI)r@rAr!r backbonerdecoder post_init)r0r!rGs rrAz VitMatteForImageMatting.__init__sX     %f-- 26::  rNr"output_attentionsoutput_hidden_stateslabels return_dictcv||n |jj}||n |jj}||n |jj}d}|t d|j|||}|jd}|||} |s| f|ddz} ||f| zn| St|| |j |j S)a8 labels (`torch.LongTensor` of shape `(batch_size, height, width)`, *optional*): Ground truth image matting for computing the loss. Examples: ```python >>> from transformers import VitMatteImageProcessor, VitMatteForImageMatting >>> import torch >>> from PIL import Image >>> from huggingface_hub import hf_hub_download >>> processor = VitMatteImageProcessor.from_pretrained("hustvl/vitmatte-small-composition-1k") >>> model = VitMatteForImageMatting.from_pretrained("hustvl/vitmatte-small-composition-1k") >>> filepath = hf_hub_download( ... repo_id="hf-internal-testing/image-matting-fixtures", filename="image.png", repo_type="dataset" ... ) >>> image = Image.open(filepath).convert("RGB") >>> filepath = hf_hub_download( ... repo_id="hf-internal-testing/image-matting-fixtures", filename="trimap.png", repo_type="dataset" ... ) >>> trimap = Image.open(filepath).convert("L") >>> # prepare image + trimap for the model >>> inputs = processor(images=image, trimaps=trimap, return_tensors="pt") >>> with torch.no_grad(): ... alphas = model(**inputs).alphas >>> print(alphas.shape) torch.Size([1, 1, 640, 960]) ```NzTraining is not yet supported)rrrzr )rrrr) r!use_return_dictrrNotImplementedErrorrforward_with_filtered_kwargs feature_mapsrrrr) r0r"rrrrroutputsrsroutputs rrLzVitMatteForImageMatting.forwards R&1%B 'tn B & B d^ B B B ^B B B B B rr)r dataclassesrtypingrrrmodeling_utilsrutilsrr utils.backbone_utilsr configuration_vitmatter rr r5r7rQrhrxrr__all__rrrrsq!!!!!! ------00000000111111222222  :::::::   :$ ) ) ) ) )o ) ) )294        F")"290&&&&&")&&&R N N N N N 5N N  N b %&? @r