o V hT< @s0ddlmZddlmZmZmZmZddlZddlm Z ddlm Z ddl m Z ddl mZdd lmZmZmZdd lmZdd lmZmZgd Zd e dede fddZGddde jZGddde jZdeededededef ddZdeddZ Gdd d eZ!Gd!d"d"eZ"Gd#d$d$eZ#Gd%d&d&eZ$eed'e!j%fd(dd)d*dee!dededefd+d,Z&eed'e"j%fd(dd)d*dee"dededefd-d.Z'eed'e#j%fd(dd)d*dee#dededefd/d0Z(eed'e$j%fd(dd)d*dee$dededefd1d2Z)dS)3)partial)AnyCallableListOptionalN)Tensor)ImageClassification)_log_api_usage_once)register_modelWeights WeightsEnum)_IMAGENET_CATEGORIES)_ovewrite_named_paramhandle_legacy_interface) ShuffleNetV2ShuffleNet_V2_X0_5_WeightsShuffleNet_V2_X1_0_WeightsShuffleNet_V2_X1_5_WeightsShuffleNet_V2_X2_0_Weightsshufflenet_v2_x0_5shufflenet_v2_x1_0shufflenet_v2_x1_5shufflenet_v2_x2_0xgroupsreturncCsP|\}}}}||}||||||}t|dd}|||||}|S)Nr r)sizeviewtorch transpose contiguous)rr batchsize num_channelsheightwidthchannels_per_groupr(S/var/www/vscode/kcb/lib/python3.10/site-packages/torchvision/models/shufflenetv2.pychannel_shuffles r*cspeZdZdedededdffdd Ze dd ed ed ededededejfddZ de de fddZ Z S)InvertedResidualinpoupstriderNcsTtd|krdkstdtd||_|d}|jdkr7||d>kr7td|d|d|d|jdkrct|j||d|jdd t|tj||ddd d d t|tj d d|_ nt|_ ttj|jdkrs|n||ddd d d t|tj d d|j||d|jdd t|tj||ddd d d t|tj d d|_ dS)Nr zillegal stride valuerzInvalid combination of stride z, inp z and oup zB values. If stride == 1 then inp should be equal to oup // 2 << 1. kernel_sizer.paddingrF)r1r.r2biasTinplace) super__init__ ValueErrorr.nn Sequentialdepthwise_conv BatchNorm2dConv2dReLUbranch1branch2)selfr,r-r.branch_features __class__r(r)r7,sJ       zInvertedResidual.__init__r rFior1r2r3c Cstj|||||||dS)N)r3r)r9r=)rErFr1r.r2r3r(r(r)r;VszInvertedResidual.depthwise_convrcCsb|jdkr|jddd\}}tj|||fdd}ntj||||fdd}t|d}|S)Nr r)dim)r.chunkr catr@r?r*)rArx1x2outr(r(r)forward\s  zInvertedResidual.forward)r rF) __name__ __module__ __qualname__intr7 staticmethodboolr9r=r;rrM __classcell__r(r(rCr)r++s(* r+c sneZdZdefdeedeedededejfddf fd d Z d e de fd d Z d e de fddZ Z S)ristages_repeatsstages_out_channels num_classesinverted_residual.rNc sdtt|t|dkrtdt|dkrtd||_d}|jd}ttj||ddddd t |tj d d |_ |}tj dddd |_ |||d ddD}t|||jddD])\}} }|||dg} t| dD] } | |||dqtt||tj| |}qb|jd}ttj||ddddd t |tj d d |_t|||_dS)Nr/z2expected stages_repeats as list of 3 positive intsz7expected stages_out_channels as list of 5 positive intsrrr F)r3Tr4r0cSsg|]}d|qS)stager().0rEr(r(r) sz)ShuffleNetV2.__init__..)rr/)r6r7r lenr8_stage_out_channelsr9r:r=r<r>conv1 MaxPool2dmaxpoolziprangeappendsetattrconv5Linearfc) rArUrVrWrXinput_channelsoutput_channels stage_namesnamerepeatsseqrErCr(r)r7isB        zShuffleNetV2.__init__rcCsX||}||}||}||}||}||}|ddg}||}|S)Nrr/)rarcstage2stage3stage4rhmeanrjrArr(r(r) _forward_impls       zShuffleNetV2._forward_implcCs ||S)N)rvrur(r(r)rMs zShuffleNetV2.forward)rNrOrPr+rrQrr9Moduler7rrvrMrTr(r(rCr)rhs  0 rweightsprogressargskwargscOsL|durt|dt|jdt|i|}|dur$||j|dd|S)NrW categoriesT)ry check_hash)rr_metarload_state_dictget_state_dict)rxryrzr{modelr(r(r) _shufflenetv2s r)r r z2https://github.com/ericsun99/Shufflenet-v2-Pytorch)min_sizer|recipec @DeZdZedeeddiedddddid d d d d ZeZdS)rzDhttps://download.pytorch.org/models/shufflenetv2_x0.5-f707e7126e.pth crop_sizei ImageNet-1Kg-FN@g9voT@zacc@1zacc@5g{Gz?gT㥛 @VThese weights were trained from scratch to reproduce closely the results of the paper. num_params_metrics_ops _file_size_docsurl transformsr~N rNrOrPr rr _COMMON_META IMAGENET1K_V1DEFAULTr(r(r(r)r& rc @r)rzBhttps://download.pytorch.org/models/shufflenetv2_x1-5666bf0f80.pthrri"rgI +WQ@gNbX9V@rg(\?gE!@rrrNrr(r(r(r)rrrc @HeZdZedeedddieddddd d id d d ddZeZdS)rzBhttps://download.pytorch.org/models/shufflenetv2_x1_5-3c479a10.pthrr resize_size+https://github.com/pytorch/vision/pull/5906iv5rg9v?R@g/$V@rgl?gw/+@ These weights were trained from scratch by using TorchVision's `new training recipe `_. rrrrrrrNrr(r(r(r)r( rc @r)rzBhttps://download.pytorch.org/models/shufflenetv2_x2_0-8be3c8ee.pthrrrriprgQS@gMb@W@rg-?g+n<@rrrNrr(r(r(r)rrr pretrained)rxT)rxrycK(t|}t||gdgdfi|S)a  Constructs a ShuffleNetV2 architecture with 0.5x output channels, as described in `ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design `__. Args: weights (:class:`~torchvision.models.ShuffleNet_V2_X0_5_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.ShuffleNet_V2_X0_5_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.shufflenetv2.ShuffleNetV2`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.ShuffleNet_V2_X0_5_Weights :members: r]r])0`)rverifyrrxryr{r(r(r)r rcKr)a  Constructs a ShuffleNetV2 architecture with 1.0x output channels, as described in `ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design `__. Args: weights (:class:`~torchvision.models.ShuffleNet_V2_X1_0_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.ShuffleNet_V2_X1_0_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.shufflenetv2.ShuffleNetV2`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.ShuffleNet_V2_X1_0_Weights :members: r)rtrir)rrrrr(r(r)r>rrcKr)a  Constructs a ShuffleNetV2 architecture with 1.5x output channels, as described in `ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design `__. Args: weights (:class:`~torchvision.models.ShuffleNet_V2_X1_5_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.ShuffleNet_V2_X1_5_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.shufflenetv2.ShuffleNetV2`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.ShuffleNet_V2_X1_5_Weights :members: r)ri`ir)rrrrr(r(r)r]rrcKr)a  Constructs a ShuffleNetV2 architecture with 2.0x output channels, as described in `ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design `__. Args: weights (:class:`~torchvision.models.ShuffleNet_V2_X2_0_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.ShuffleNet_V2_X2_0_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.shufflenetv2.ShuffleNetV2`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.ShuffleNet_V2_X2_0_Weights :members: r)riii)rrrrr(r(r)r|rr)* functoolsrtypingrrrrr torch.nnr9rtransforms._presetsr utilsr _apir r r_metar_utilsrr__all__rQr*rwr+rrSrrrrrrrrrrrr(r(r(r)s       =A