ڧg5ddlZddlmZmZmZmZmZmZddlZddlm Z ddl m Z m Z ej jjZGddej jZGdd ej jZGd d eZGd d eZGddej jZGddej jZGddej jZdS)N)CallableListOptionalSequenceTupleUnion)Tensor)_log_api_usage_once _make_ntupleceZdZdZ ddedeffd Zdededed e d e ed e ed e effd Z de de fdZ defdZxZS)FrozenBatchNorm2da! BatchNorm2d where the batch statistics and the affine parameters are fixed Args: num_features (int): Number of features ``C`` from an expected input of size ``(N, C, H, W)`` eps (float): a value added to the denominator for numerical stability. Default: 1e-5 h㈵> num_featuresepsctt|||_|dt j||dt j||dt j||dt j|dS)Nweightbias running_mean running_var)super__init__r rregister_buffertorchoneszeros)selfrr __class__s P/var/www/html/ai-engine/env/lib/python3.11/site-packages/torchvision/ops/misc.pyrzFrozenBatchNorm2d.__init__s D!!! Xuz,'?'?@@@ VU[%>%>??? ^U[-F-FGGG ]EJ|,D,DEEEEE state_dictprefixlocal_metadatastrict missing_keysunexpected_keys error_msgsc n|dz}||vr||=t|||||||dS)Nnum_batches_tracked)r_load_from_state_dict) rr!r"r#r$r%r&r'num_batches_tracked_keyrs rr*z'FrozenBatchNorm2d._load_from_state_dict#s[#)+@"@ "j 0 023 %%  oWa     r xreturncH|jdddd}|jdddd}|jdddd}|jdddd}|||jzz}|||zz }||z|zS)N)rreshaperrrrrsqrt)rr,wbrvrmscalers rforwardzFrozenBatchNorm2d.forward5s K  2q! , , I  aQ * *   % %aQ 2 2   & &q"a 3 3R$(]))+++2:~5y4r cV|jjd|jjdd|jdS)N(rz, eps=))r__name__rshaper)rs r__repr__zFrozenBatchNorm2d.__repr__@s2.)SSDK,=a,@SSSSSSr )r)r< __module__ __qualname____doc__intfloatrdictstrboolrr*r r8r> __classcell__rs@rrr s  F F F F F F F F F       3i  c I      $    F     T#TTTTTTTTr rceZdZddddejjejjdddejjf dedede ee edffd e ee edffd e e ee edfe fd ed e e dejjfd e e dejjfde ee edffde ede ede dejjfddffd ZxZS)ConvNormActivationr/NT in_channels out_channels kernel_size.stridepaddinggroups norm_layeractivation_layerdilationinplacer conv_layerr-c  |ttr!t tr dz dz z}n|ttrtnt } t | t |  t fdt | D}| |du} | |||| || g}|||||#| ind| i}||di|tj |t|||_ |j tkrtjddSdS)Nr/r c3FK|]}|dz dz|zVdS)r/r N).0irTrNs r z.ConvNormActivation.__init__..\s:bbAQ!!3 9HQK Gbbbbbbr )rTrQrrUzhDon't use ConvNormActivation directly, please use Conv2dNormActivation and Conv3dNormActivation instead.rY) isinstancerBrlenr tuplerangeappendrrr rMrrJwarningswarn)rrLrMrNrOrPrQrRrSrTrUrrV _conv_dimlayersparamsrs ` ` rrzConvNormActivation.__init__Es ?+s++ c 8S0I0I c&?q08;0:;0Q0QdC ,,,WZ[cWdWd *; BB ')<<bbbbbQVW`QaQabbbbb <%D J!       ! MM**\22 3 3 3  '"?RRG0DF MM**44V44 5 5 5&!!D!!!( >/ / / Mz      0 /r )r<r?r@rnn BatchNorm2dReLUConv2drBrrrrErModulerFrrGrHs@rrJrJDsz 45./>B?Dx?SEJX]01"&#5:X_55553c3h/0 5 c5c?*+ 5 %U38_c 9:; 55Xc58?&:;<5#8C,@#AB5U38_,-5$5tn5S%(/125 5555555555r rJceZdZdZddddejjejjdddf dedede ee eeffd e ee eeffd e e ee eefe fd ed e e d ejjfde e d ejjfde ee eeffde ede eddffd ZxZS)Conv2dNormActivationa Configurable block used for Convolution2d-Normalization-Activation blocks. Args: in_channels (int): Number of channels in the input image out_channels (int): Number of channels produced by the Convolution-Normalization-Activation block kernel_size: (int, optional): Size of the convolving kernel. Default: 3 stride (int, optional): Stride of the convolution. Default: 1 padding (int, tuple or str, optional): Padding added to all four sides of the input. Default: None, in which case it will be calculated as ``padding = (kernel_size - 1) // 2 * dilation`` groups (int, optional): Number of blocked connections from input channels to output channels. Default: 1 norm_layer (Callable[..., torch.nn.Module], optional): Norm layer that will be stacked on top of the convolution layer. If ``None`` this layer won't be used. Default: ``torch.nn.BatchNorm2d`` activation_layer (Callable[..., torch.nn.Module], optional): Activation function which will be stacked on top of the normalization layer (if not None), otherwise on top of the conv layer. If ``None`` this layer won't be used. Default: ``torch.nn.ReLU`` dilation (int): Spacing between kernel elements. Default: 1 inplace (bool): Parameter for the activation layer, which can optionally do the operation in-place. Default ``True`` bias (bool, optional): Whether to use bias in the convolution layer. By default, biases are included if ``norm_layer is None``. rKr/NTrLrMrNrOrPrQrR.rSrTrUrr-c ~t||||||||| | | tjj dSN)rrrrgrj rrLrMrNrOrPrQrRrSrTrUrrs rrzConv2dNormActivation.__init__S             HO r )r<r?r@rArrgrhrirBrrrrErrkrFrrGrHs@rrmrm}se,45./>B?Dx?SEJX]01"&#    3c3h/0  c5c?*+  %U38_c 9:;   Xc58?&:;< #8C,@#AB U38_,- $ tn            r rmceZdZdZddddejjejjdddf dedede ee eeeffd e ee eeeffd e e ee eeefe fd ed e e d ejjfde e d ejjfde ee eeeffde ede eddffd ZxZS)Conv3dNormActivationa Configurable block used for Convolution3d-Normalization-Activation blocks. Args: in_channels (int): Number of channels in the input video. out_channels (int): Number of channels produced by the Convolution-Normalization-Activation block kernel_size: (int, optional): Size of the convolving kernel. Default: 3 stride (int, optional): Stride of the convolution. Default: 1 padding (int, tuple or str, optional): Padding added to all four sides of the input. Default: None, in which case it will be calculated as ``padding = (kernel_size - 1) // 2 * dilation`` groups (int, optional): Number of blocked connections from input channels to output channels. Default: 1 norm_layer (Callable[..., torch.nn.Module], optional): Norm layer that will be stacked on top of the convolution layer. If ``None`` this layer won't be used. Default: ``torch.nn.BatchNorm3d`` activation_layer (Callable[..., torch.nn.Module], optional): Activation function which will be stacked on top of the normalization layer (if not None), otherwise on top of the conv layer. If ``None`` this layer won't be used. Default: ``torch.nn.ReLU`` dilation (int): Spacing between kernel elements. Default: 1 inplace (bool): Parameter for the activation layer, which can optionally do the operation in-place. Default ``True`` bias (bool, optional): Whether to use bias in the convolution layer. By default, biases are included if ``norm_layer is None``. rKr/NTrLrMrNrOrPrQrR.rSrTrUrr-c ~t||||||||| | | tjj dSro)rrrrgConv3drps rrzConv3dNormActivation.__init__rqr )r<r?r@rArrg BatchNorm3drirBrrrrErrkrFrrGrHs@rrsrssq*9:34CG?Dx?SEJX]56"&#    3c3m 445  c5c3//0  %U3S=%93 >?@   Xc58?&:;< #8C,@#AB U3S=112 $ tn            r rsc eZdZdZejjejjfdedede dejj fde dejj fddf fd Z d e de fd Z d e de fd ZxZS) SqueezeExcitationaE This block implements the Squeeze-and-Excitation block from https://arxiv.org/abs/1709.01507 (see Fig. 1). Parameters ``activation``, and ``scale_activation`` correspond to ``delta`` and ``sigma`` in eq. 3. Args: input_channels (int): Number of channels in the input image squeeze_channels (int): Number of squeeze channels activation (Callable[..., torch.nn.Module], optional): ``delta`` activation. Default: ``torch.nn.ReLU`` scale_activation (Callable[..., torch.nn.Module]): ``sigma`` activation. Default: ``torch.nn.Sigmoid`` input_channelssqueeze_channels activation.scale_activationr-Nctt|tjd|_tj||d|_tj||d|_ ||_ ||_ dS)Nr/) rrr rrgAdaptiveAvgPool2davgpoolrjfc1fc2r{r|)rryrzr{r|rs rrzSqueezeExcitation.__init__s D!!!x11!44 8??>3CQGG8??#3^QGG$*,, 0 0 2 2r inputc||}||}||}||}||Sro)rrr{rr|rrr7s r_scalezSqueezeExcitation._scalesV U##&&$$U+++r c6||}||zSro)rrs rr8zSqueezeExcitation.forwards E""u}r )r<r?r@rArrgriSigmoidrBrrkrr rr8rGrHs@rrxrxs  6;X];@8;K 3 3 3 3S%(/12 3 #3#78 3  3 3 3 3 3 3,F,v,,,,Vr rxceZdZdZdejjdddfdedeede e dejj fd e e dejj fd e e d e d e ffd ZxZS)MLPaThis block implements the multi-layer perceptron (MLP) module. Args: in_channels (int): Number of channels of the input hidden_channels (List[int]): List of the hidden channel dimensions norm_layer (Callable[..., torch.nn.Module], optional): Norm layer that will be stacked on top of the linear layer. If ``None`` this layer won't be used. Default: ``None`` activation_layer (Callable[..., torch.nn.Module], optional): Activation function which will be stacked on top of the normalization layer (if not None), otherwise on top of the linear layer. If ``None`` this layer won't be used. Default: ``torch.nn.ReLU`` inplace (bool, optional): Parameter for the activation layer, which can optionally do the operation in-place. Default is ``None``, which uses the respective default values of the ``activation_layer`` and Dropout layer. bias (bool): Whether to use bias in the linear layer. Default ``True`` dropout (float): The probability for the dropout layer. Default: 0.0 NTgrLhidden_channelsrR.rSrUrdropoutc|ind|i}g} |} |ddD]} | tj| | ||| || | |di|| tjj|fi|| } | tj| |d|| tjj|fi|t j| t|dS)NrUr0)rrY)rarrgLinearDropoutrrr ) rrLrrRrSrUrrrfrein_dim hidden_dimrs rrz MLP.__init__sMY,@)#2#.  J MM%(//&*4/HH I I I% jj44555 MM**44V44 5 5 5 MM%(*7==f== > > >FF ehoofob.AoMMNNN eh&w99&99:::&!!D!!!!!r )r<r?r@rArrgrirBrrrrkrFrCrrGrHs@rrrs  "@DEJX]"&"""c"Xc58?&:;< " #8C,@#AB " $ """"""""""""r rcDeZdZdZdeeffd ZdedefdZxZ S)PermutezThis module returns a view of the tensor input with its dimensions permuted. Args: dims (List[int]): The desired ordering of dimensions dimscVt||_dSro)rrr)rrrs rrzPermute.__init__;s$  r r,r-c6tj||jSro)rpermuter)rr,s rr8zPermute.forward?s}Q ***r ) r<r?r@rArrBrr r8rGrHs@rrr4st T#Y++F++++++++r r)rbtypingrrrrrrrr utilsr r rg functional interpolaterkr SequentialrJrmrsrxrrrYr rrsCCCCCCCCCCCCCCCC 55555555h!- 4T4T4T4T4T4T4T4Tn66666,666r/ / / / / -/ / / d. . . . . -. . . b$$$$$$$$N*"*"*"*"*"%( *"*"*"Z + + + + +eho + + + + +r