o ,h? @sddlmZddlmZddlmZmZmZddlZddlm Z m Z ddl m Z m ZddlmZdd lmZd d lmZmZmZd d lmZd d lmZmZmZgdZGdddZGddde jZ Gddde jZ! d3de"de#de$de$def ddZ%de&ede'd eed!e$ded"e!f d#d$Z(d%ed&Z)Gd'd(d(eZ*Gd)d*d*eZ+eed+e*j,fd,dd-d.d ee*d!e$ded"e!fd/d0Z-eed+e+j,fd,dd-d.d ee+d!e$ded"e!fd1d2Z.dS)4)Sequence)partial)AnyCallableOptionalN)nnTensor)Conv2dNormActivationSqueezeExcitation)ImageClassification)_log_api_usage_once)register_modelWeights WeightsEnum)_IMAGENET_CATEGORIES)_make_divisible_ovewrite_named_paramhandle_legacy_interface) MobileNetV3MobileNet_V3_Large_WeightsMobileNet_V3_Small_Weightsmobilenet_v3_largemobilenet_v3_smallc@sPeZdZdedededededededed efd d Zed ed efd dZ dS)InvertedResidualConfiginput_channelskernelexpanded_channels out_channelsuse_se activationstridedilation width_multc CsP||| |_||_||| |_||| |_||_|dk|_||_||_dS)NHS) adjust_channelsrrrrr use_hsr"r#) selfrrrrr r!r"r#r$r)Y/var/www/html/scripts/venv/lib/python3.10/site-packages/torchvision/models/mobilenetv3.py__init__s   zInvertedResidualConfig.__init__channelscCst||dS)N)r)r,r$r)r)r*r&0sz&InvertedResidualConfig.adjust_channelsN) __name__ __module__ __qualname__intboolstrfloatr+ staticmethodr&r)r)r)r*rs,  rc s^eZdZeeejdfdededej fdedej fffdd Z de d e fd d Z Z S) InvertedResidual)scale_activationcnf norm_layer.se_layerc s*td|jkrdkstdtd|jdko"|j|jk|_g}|jr,tj ntj }|j |jkrC| t |j|j d||d|jdkrJdn|j}| t |j |j |j||j|j ||d|jrut|j dd}| ||j || t |j |jd|ddtj||_|j|_|jdk|_dS)Nrr zillegal stride value kernel_sizer9activation_layer)r<r"r#groupsr9r=r-)superr+r" ValueErrorrruse_res_connectr'r HardswishReLUrappendr r#rr r Sequentialblock_is_cn)r(r8r9r:layersr=r"squeeze_channels __class__r)r*r+7sT     zInvertedResidual.__init__inputreturncCs||}|jr ||7}|SN)rGrB)r(rMresultr)r)r*forwardos zInvertedResidual.forward)r.r/r0rSElayerr HardsigmoidrrModuler+rrQ __classcell__r)r)rKr*r65s   8r6cseZdZ    ddeedededeedej fd eedej fd e d e d dffd d Z de d e fddZde d e fddZZS)rN皙?inverted_residual_setting last_channel num_classesrG.r9dropoutkwargsrNc stt||stdt|trtdd|Ds!td|dur't}|dur3t t j ddd}g}|d j } | td | d d |t jd |D] } | || |qJ|d j} d| } | t| | d|t jdt j||_t d|_t t | |t jddt j|ddt |||_|D]L} t| t jrt jj| jdd| jdurt j| jqt| t j t jfrt j | jt j| jqt| t jrt j!| jd dt j| jqdS)a. MobileNet V3 main class Args: inverted_residual_setting (List[InvertedResidualConfig]): Network structure last_channel (int): The number of channels on the penultimate layer num_classes (int): Number of classes block (Optional[Callable[..., nn.Module]]): Module specifying inverted residual building block for mobilenet norm_layer (Optional[Callable[..., nn.Module]]): Module specifying the normalization layer to use dropout (float): The droupout probability z1The inverted_residual_setting should not be emptycSsg|]}t|tqSr)) isinstancer).0sr)r)r* sz(MobileNetV3.__init__..zDThe inverted_residual_setting should be List[InvertedResidualConfig]NgMbP?g{Gz?)epsmomentumrr )r<r"r9r=rr;T)inplace)prffan_out)mode)"r@r+r rAr]rall TypeErrorr6rr BatchNorm2drrEr rCrrFfeaturesAdaptiveAvgPool2davgpoolLinearDropout classifiermodulesConv2dinitkaiming_normal_weightbiaszeros_ GroupNormones_normal_)r(rXrYrZrGr9r[r\rIfirstconv_output_channelsr8lastconv_input_channelslastconv_output_channelsmrKr)r*r+wsx             zMobileNetV3.__init__xcCs.||}||}t|d}||}|S)Nr)rmrotorchflattenrrr(rr)r)r* _forward_impls    zMobileNetV3._forward_implcCs ||SrO)rrr)r)r*rQs zMobileNetV3.forward)rVNNrW)r.r/r0listrr1rrrrTr4rr+rrrQrUr)r)rKr*rvs0 [ r?Farchr$ reduced_taildilatedr\c Ks|rdnd}|r dnd}tt|d}ttj|d}|dkr|dddddddd|ddd d dddd|d dd d dddd|d d d d dddd|d d dd dddd|d d dd dddd|d ddddddd|dddddddd|dddddddd|dddddddd|dddddddd|dddddddd|dd dd|ddd||d|d d|d|ddd||d|d d|d|ddd|g} |d|} | | fS|dkrZ|dddddddd|ddd d dddd|d ddd dddd|d d dd dddd|d d dd dddd|d d dd dddd|d d dddddd|dd dddddd|dd d d|ddd||d|d d!|d|ddd||d|d d!|d|ddd|g } |d"|} | | fStd#|)$Nr r)r$rrcFRE@H(TxPr%ipiiirX`0i i@izUnsupported model type )rrr&rA) rr$rrr\reduce_dividerr# bneck_confr&rXrYr)r)r*_mobilenet_v3_confsN          rrXrYweightsprogressrNcKsP|durt|dt|jdt||fi|}|dur&||j|dd|S)NrZ categoriesT)r check_hash)rlenmetarload_state_dictget_state_dict)rXrYrrr\modelr)r)r* _mobilenet_v3s r)rr)min_sizerc @s~eZdZedeeddiedddddd id d d d dZedeedddiedddddd id ddd dZeZ dS)rzChttps://download.pytorch.org/models/mobilenet_v3_large-8738ca79.pth crop_sizeiS^https://github.com/pytorch/vision/tree/main/references/classification#mobilenetv3-large--small ImageNet-1Kg R@g(\V@zacc@1zacc@5g-?gw/5@zJThese weights were trained from scratch by using a simple training recipe. num_paramsrecipe_metrics_ops _file_size_docsurl transformsrzChttps://download.pytorch.org/models/mobilenet_v3_large-5c1a4163.pth)r resize_sizezHhttps://github.com/pytorch/vision/issues/3995#new-recipe-with-reg-tuninggK7R@gNbX9$W@gZd5@a/ These weights improve marginally upon the results of the original paper by using a modified version of TorchVision's `new training recipe `_. N) r.r/r0rrr _COMMON_META IMAGENET1K_V1 IMAGENET1K_V2DEFAULTr)r)r)r*r)sL  rc @sFeZdZedeeddiedddddd id d d d dZeZdS)rzChttps://download.pytorch.org/models/mobilenet_v3_small-047dcff4.pthrri&rrgnP@g}?5^U@rgv/?g r#@z} These weights improve upon the results of the original paper by using a simple training recipe. rrN) r.r/r0rrr rrrr)r)r)r*rUs( r pretrained)rT)rrcK2t|}tdi|\}}t||||fi|S)a Constructs a large MobileNetV3 architecture from `Searching for MobileNetV3 `__. Args: weights (:class:`~torchvision.models.MobileNet_V3_Large_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.MobileNet_V3_Large_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.mobilenet.MobileNetV3`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.MobileNet_V3_Large_Weights :members: rN)r)rverifyrrrrr\rXrYr)r)r*rm rcKr)a Constructs a small MobileNetV3 architecture from `Searching for MobileNetV3 `__. Args: weights (:class:`~torchvision.models.MobileNet_V3_Small_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.MobileNet_V3_Small_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.mobilenet.MobileNetV3`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.MobileNet_V3_Small_Weights :members: rN)r)rrrrrr)r)r*rrr)rFF)/collections.abcr functoolsrtypingrrrrrrops.miscr r rRtransforms._presetsr utilsr _apirrr_metar_utilsrrr__all__rrTr6rr3r4r2rrr1rrrrrrrr)r)r)r*s     Ak 1 ,