o V h5@sddlZddlmZmZmZmZmZmZddlZddlm Z ddl m Z m Z ej jjZGdddej jZGdd d ej jZGd d d eZGd d d eZGdddej jZGdddej jZGdddej jZdS)N)CallableListOptionalSequenceTupleUnion)Tensor)_log_api_usage_once _make_ntuplecseZdZdZ ddedeffdd Zdeded ed e d e ed e ed e effdd Z de de fddZ defddZZS)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_featuresepscsdtt|||_|dt||dt||dt||dt|dS)Nweightbias running_mean running_var)super__init__r rregister_buffertorchoneszeros)selfrr __class__H/var/www/vscode/kcb/lib/python3.10/site-packages/torchvision/ops/misc.pyrs zFrozenBatchNorm2d.__init__ state_dictprefixlocal_metadatastrict missing_keysunexpected_keys error_msgsc s2|d}||vr ||=t|||||||dS)Nnum_batches_tracked)r_load_from_state_dict) rrr r!r"r#r$r%num_batches_tracked_keyrrrr'#s  z'FrozenBatchNorm2d._load_from_state_dictxreturncCsr|jdddd}|jdddd}|jdddd}|jdddd}|||j}|||}|||S)N)rreshaperrrrrsqrt)rr)wbrvrmscalerrrrforward5s  zFrozenBatchNorm2d.forwardcCs$|jjd|jjdd|jdS)N(rz, eps=))r__name__rshaper)rrrr__repr__@s$zFrozenBatchNorm2d.__repr__)r )r7 __module__ __qualname____doc__intfloatrdictstrboolrr'rr4r9 __classcell__rrrrr s2   r cseZdZddddejjejjdddejjf dedede ee edffd e ee edffd e e ee edfe fd ed e e dejjfd e e dejjfde ee edffde ede ede dejjfddffdd ZZS)ConvNormActivationr+NT in_channels out_channels kernel_size.stridepaddinggroups norm_layeractivation_layerdilationinplacer conv_layerr*c  s|dur\s&z.ConvNormActivation.__init__..)rMrJrrNzhDon't use ConvNormActivation directly, please use Conv2dNormActivation and Conv3dNormActivation instead.r) isinstancer=rlenr tuplerangeappendrrr rFrrCwarningswarn)rrErFrGrHrIrJrKrLrMrNrrO _conv_dimlayersparamsrrRrrEsB     zConvNormActivation.__init__)r7r:r;rnn BatchNorm2dReLUConv2dr=rrrr@rModulerArrBrrrrrCDsL     rCcseZdZdZddddejjejjdddf dedede ee eeffd e ee eeffd e e ee eefe fd ed e e d ejjfde e d ejjfde ee eeffde ede eddffdd ZZS)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``. rDr+NTrErFrGrHrIrJrK.rLrMrNrr*c *t||||||||| | | tjj dSN)rrrr^ra rrErFrGrHrIrJrKrLrMrNrrrrrzConv2dNormActivation.__init__)r7r:r;r<rr^r_r`r=rrrr@rrbrArrBrrrrrc}sH     rccseZdZdZddddejjejjdddf dedede ee eeeffd e ee eeeffd e e ee eeefe fd ed e e d ejjfde e d ejjfde ee eeeffde ede eddffdd ZZS)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``. rDr+NTrErFrGrHrIrJrK.rLrMrNrr*c rdre)rrrr^Conv3drfrrrrrgzConv3dNormActivation.__init__)r7r:r;r<rr^ BatchNorm3dr`r=rrrr@rrbrArrBrrrrrhsH     rhc seZdZdZejjejjfdedede dejj fde dejj fddf fd d Z d e de fd d Z d e de fddZZS)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*NcsXtt|tjd|_tj||d|_tj||d|_ ||_ ||_ dS)Nr+) rrr rr^AdaptiveAvgPool2davgpoolrafc1fc2rnro)rrlrmrnrorrrrs  zSqueezeExcitation.__init__inputcCs2||}||}||}||}||Sre)rqrrrnrsrorrtr3rrr_scales     zSqueezeExcitation._scalecCs||}||Sre)rvrurrrr4s zSqueezeExcitation.forward)r7r:r;r<rr^r`Sigmoidr=rrbrrrvr4rBrrrrrks"rkcsveZdZdZdejjdddfdedeede e dejj fd e e dejj fd e e d e d e ffd d ZZS)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 NTgrEhidden_channelsrK.rLrNrdropoutc s|durind|i}g} |} |ddD]2} | tjj| | |d|dur-| || | |di|| tjj|fi|| } q| tjj| |d|d| tjj|fi|tj| t|dS)NrNr,)rr)rXrr^LinearDropoutrrr ) rrEryrKrLrNrrzr]r\in_dim hidden_dimrrrrs   z MLP.__init__)r7r:r;r<rr^r`r=rrrrbrAr>rrBrrrrrxs,rxcs<eZdZdZdeeffdd ZdedefddZZ S) PermutezThis module returns a view of the tensor input with its dimensions permuted. Args: dims (List[int]): The desired ordering of dimensions dimscst||_dSre)rrr)rrrrrr;s  zPermute.__init__r)r*cCst||jSre)rpermuter)rr)rrrr4?szPermute.forward) r7r:r;r<rr=rrr4rBrrrrr4sr)rYtypingrrrrrrrrutilsr r r^ functional interpolaterbr SequentialrCrcrhrkrxrrrrrs   7921'-