Ei5SSKrSSKJr SSKJrJrJr SSKrSSKJr SSK J r J r \RRRr"SS\RR5r"S S \RR"5r"S S \5r"S S\5r"SS\RR5r"SS\RR"5r"SS\RR5rg)N)Sequence)CallableOptionalUnion)Tensor)_log_api_usage_once _make_ntuplec^\rSrSrSrSS\S\4U4SjjjrS\S\ S\S \ S \ \ S \ \ S \ \ 4U4S jjr S\ S\ 4SjrS\ 4SjrSrU=r$)FrozenBatchNorm2da  BatchNorm2d where the batch statistics and the affine parameters are fixed Args: num_features (int): Number of features ``C`` from an expected input of size ``(N, C, H, W)`` eps (float): a value added to the denominator for numerical stability. Default: 1e-5 num_featuresepsct>[TU]5 [U5 X lUR S[ R "U55 UR S[ R"U55 UR S[ R"U55 UR S[ R "U55 g)Nweightbias running_mean running_var)super__init__r rregister_buffertorchoneszeros)selfrr __class__s Q/home/mande/repo/quber/.venv/lib/python3.13/site-packages/torchvision/ops/misc.pyrFrozenBatchNorm2d.__init__s D! Xuzz,'?@ VU[[%>? ^U[[-FG ]EJJ|,DE state_dictprefixlocal_metadatastrict missing_keysunexpected_keys error_msgsc B>US-nX;aX [T U]XX4XVU5 g)Nnum_batches_tracked)r_load_from_state_dict) rr r!r"r#r$r%r&num_batches_tracked_keyrs rr)'FrozenBatchNorm2d._load_from_state_dict$s4#)+@"@ " 03 %  Wa rxreturncLURRSSSS5nURRSSSS5nURRSSSS5nURRSSSS5nX$UR -R 5-nX5U-- nX-U-$)N)rreshaperrrrrsqrt)rr,wbrvrmscalers rforwardFrozenBatchNorm2d.forward6s KK  2q! , II  aQ *    % %aQ 2    & &q"a 3$((]))++:~y4rcURRSURRSSURS3$)N(rz, eps=))r__name__rshaper)rs r__repr__FrozenBatchNorm2d.__repr__As;..))*!DKK,=,=a,@+AzQRSSr)r)gh㈵>)r= __module__ __qualname____firstlineno____doc__intfloatrdictstrboollistr)rr8r?__static_attributes__ __classcell__rs@rr r s F F F F       3i  c I $    F  T#TTrr c^\rSrSrSSSS\R R \R RSSS\R R4 S\ S\ S\ \ \ \ S 44S \ \ \ \ S 44S \ \ \ \ \ S 4\ 4S \ S \ \S \R R4S\ \S \R R4S\ \ \ \ S 44S\ \S\ \S\S \R R4SS4U4SjjjrSrU=r$)ConvNormActivationEr/NT in_channels out_channels kernel_size.stridepaddinggroups norm_layeractivation_layerdilationinplacer conv_layerr-c v>^^ Uc[T[5(a![T [5(a TS- S-T -nOb[T[5(a [T5O [T 5n [ TU 5m[ T U 5m [ U U4Sj[ U 555nU cUSLn U "UUTUUT UU S9/nUbURU"U55 Ub U c0OSU 0nURU"S0UD65 [TU]$"U6 [U5 X l UR[:Xa[R"S5 gg)Nr/rc3F># UHnTUS- S-TU-v M g7f)r/rN).0irZrTs r .ConvNormActivation.__init__..]s*bQaAQ!!3 9HQK GQas!)rZrWrr[zhDon't use ConvNormActivation directly, please use Conv2dNormActivation and Conv3dNormActivation instead.r_) isinstancerErlenr tuplerangeappendrrr rSrrOwarningswarn)rrRrSrTrUrVrWrXrYrZr[rr\ _conv_dimlayersparamsrs ` ` rrConvNormActivation.__init__Fs< ?+s++ 8S0I0I&?q08;0:;0Q0QC ,WZ[cWd *; B ')<bQVW`Qabb <%D !     ! MM*\2 3  '"?RG0DF MM*4V4 5 &!D!( >>/ / MMz  0r)rS)r=rArBrCrnn BatchNorm2dReLUConv2drErrfrrHrModulerIrrKrLrMs@rrOrOEs^ 45./>B?Dxx?S?SEJXX]]01"&#5:XX__5553c3h/0 5 c5c?*+ 5 %U38_c 9:; 55Xc588??&:;<5#8C,@#AB5U38_,-5$5tn5S%((//125 55rrOc^\rSrSrSrSSSS\R R\R RSSS4 S\ S\ S \ \ \ \ \ 44S \ \ \ \ \ 44S \ \ \ \ \ \ 4\ 4S \ S \ \S\R R4S\ \S\R R4S\ \ \ \ \ 44S\ \S\ \SS4U4SjjjrSrU=r$)Conv2dNormActivation~a 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``. rQr/NTrRrSrTrUrVrWrX.rYrZr[rr-c j>[T U]UUUUUUUUU U U [RR5 gN)rrrrorr rrRrSrTrUrVrWrXrYrZr[rrs rrConv2dNormActivation.__init__>             HHOO rr_)r=rArBrCrDrrorprqrErrfrrHrrsrIrrKrLrMs@rruru~s<,45./>B?Dxx?S?SEJXX]]01"&#   3c3h/0  c5c?*+  %U38_c 9:;   Xc588??&:;< #8C,@#AB U38_,- $ tn    rruc^\rSrSrSrSSSS\R R\R RSSS4 S\ S\ S \ \ \ \ \ \ 44S \ \ \ \ \ \ 44S \ \ \ \ \ \ \ 4\ 4S \ S \ \S\R R4S\ \S\R R4S\ \ \ \ \ \ 44S\ \S\ \SS4U4SjjjrSrU=r$)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``. rQr/NTrRrSrTrUrVrWrX.rYrZr[rr-c j>[T U]UUUUUUUUU U U [RR5 grx)rrrroConv3drys rrConv3dNormActivation.__init__r{rr_)r=rArBrCrDrro BatchNorm3drqrErrfrrHrrsrIrrKrLrMs@rr}r}sH*9:34CG?Dxx?S?SEJXX]]56"&#   3c3m 445  c5c3//0  %U3S=%93 >?@   Xc588??&:;< #8C,@#AB U3S=112 $ tn    rr}c $^\rSrSrSr\R R\R R4S\ S\ S\ S\R R4S\ S\R R4SS 4 U4S jjjr S \ S\ 4S jrS \ S\ 4S jrSrU=r$)SqueezeExcitationa% 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-NcD>[TU]5 [U5 [RR S5Ul[RRXS5Ul[RRX!S5Ul U"5Ul U"5Ul g)Nr/) rrr rroAdaptiveAvgPool2davgpoolrrfc1fc2rr)rrrrrrs rrSqueezeExcitation.__init__so D!xx11!4 88??>QG88??#3QG$, 0 2rinputcURU5nURU5nURU5nURU5nUR U5$rx)rrrrrrrr7s r_scaleSqueezeExcitation._scalesI U#&$$U++rc,URU5nX!-$rx)rrs rr8SqueezeExcitation.forwards E"}r)rrrrr)r=rArBrCrDrrorqSigmoidrErrsrrrr8rKrLrMs@rrrs 6;XX]];@88;K;K 3 3 3S%((//12 3 #3#78 3  3 3,F,v,Vrrc^\rSrSrSrS\R RSSS4S\S\ \S\ \ S \R R4S \ \ S \R R4S \ \ S \ S \4U4SjjjrSrU=r$)MLPiaThis 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 NTgrRhidden_channelsrX.rYr[rdropoutc H>Uc0OSU0n/n Un USSHn U R[RRXUS95 UbU RU"U 55 U RU"S0UD65 U R[RR"U40UD65 U n M U R[RRXSUS95 U R[RR"U40UD65 [ T U]"U 6 [U5 g)Nr[r0)rr_)rhrroLinearDropoutrrr ) rrRrrXrYr[rrrmrlin_dim hidden_dimrs rr MLP.__init__sY,@)#2.J MM%((//&4/H I% j45 MM*4V4 5 MM%((**7=f= >F /  ehhoofb.AoMN ehh&&w9&9: &!D!rr_)r=rArBrCrDrrorqrErJrrrsrIrFrrKrLrMs@rrrs "@DEJXX]]"&""c"Xc588??&:;< " #8C,@#AB " $ """"rrcL^\rSrSrSrS\\4U4SjjrS\S\4Sjr Sr U=r $) Permutei5zThis module returns a view of the tensor input with its dimensions permuted. Args: dims (List[int]): The desired ordering of dimensions dimsc.>[TU]5 Xlgrx)rrr)rrrs rrPermute.__init__<s  rr,r-cB[R"XR5$rx)rpermuter)rr,s rr8Permute.forward@s}}Q **r)r) r=rArBrCrDrJrErrr8rKrLrMs@rrr5s0 T#Y++F++rr)ricollections.abcrtypingrrrrrutilsr r ro functional interpolatersr SequentialrOrur}rrrr_rrrs$,, 5hh!!-- 4T4Tn6,,6r/ -/ d. -. b$$N*"%((  *"Z +ehhoo +r