o ^i@sddlZddlZddlmmZddlmZmZddlmZ dded e d e d e d e d e defddZ dded e d e d e d e d e defddZ ejdGdddejZejdGdddeZdS)N)nnTensor)_log_api_usage_onceFư>Tinputp block_sizeinplaceepstrainingreturnc Csztjstjstt|dks|dkrtd|d|jdkr,td|jd|r2|dkr4|S|\}}}} t || |}|dd krPtd |d ||| |d||d | |d } tj ||||d | |d f|j |j d } | | tj| |dgdd d} tj| d||f|dd} d | } | || } |r|| | |S|| | }|S)a Implements DropBlock2d from `"DropBlock: A regularization method for convolutional networks" `. Args: input (Tensor[N, C, H, W]): The input tensor or 4-dimensions with the first one being its batch i.e. a batch with ``N`` rows. p (float): Probability of an element to be dropped. block_size (int): Size of the block to drop. inplace (bool): If set to ``True``, will do this operation in-place. Default: ``False``. eps (float): A value added to the denominator for numerical stability. Default: 1e-6. training (bool): apply dropblock if is ``True``. Default: ``True``. Returns: Tensor[N, C, H, W]: The randomly zeroed tensor after dropblock. ?4drop probability has to be between 0 and 1, but got .z#input should be 4 dimensional. Got dimensions.rrblock size should be odd. Got  which is even.dtypedevicevalue)rrstride kernel_sizepadding)torchjit is_scripting is_tracingr drop_block2d ValueErrorndimsizeminemptyrr bernoulli_Fpad max_pool2dnumelsummul_) rrr r r r NCHWgammanoisenormalize_scaler8U/var/www/html/RAG/RAG_venv/lib/python3.10/site-packages/torchvision/ops/drop_block.pyr$ s0    ,,  r$cCstjstjstt|dks|dkrtd|d|jdkr,td|jd|r2|dkr4|S|\}}}} } t ||| | }|dd krRtd |d ||| | |d ||d | |d | |d } tj ||||d | |d | |d f|j |j d} | | tj| |dgdd d} tj| d|||f|dd} d | } | || } |r|| | |S|| | }|S)a  Implements DropBlock3d from `"DropBlock: A regularization method for convolutional networks" `. Args: input (Tensor[N, C, D, H, W]): The input tensor or 5-dimensions with the first one being its batch i.e. a batch with ``N`` rows. p (float): Probability of an element to be dropped. block_size (int): Size of the block to drop. inplace (bool): If set to ``True``, will do this operation in-place. Default: ``False``. eps (float): A value added to the denominator for numerical stability. Default: 1e-6. training (bool): apply dropblock if is ``True``. Default: ``True``. Returns: Tensor[N, C, D, H, W]: The randomly zeroed tensor after dropblock. rrrrz#input should be 5 dimensional. Got rrrrrrrr)rrrr)r r!r"r#r drop_block3dr%r&r'r(r)rrr*r+r, max_pool3dr.r/r0)rrr r r r r1r2Dr3r4r5r6r7r8r8r9r=:s8   <,  r=c sXeZdZdZddededededd f fd d Zd edefd dZ de fddZ Z S) DropBlock2dz# See :func:`drop_block2d`. Frrr r r r Ncs&t||_||_||_||_dSN)super__init__rr r r selfrr r r  __class__r8r9rCws  zDropBlock2d.__init__rcCt||j|j|j|j|jSz Args: input (Tensor): Input feature map on which some areas will be randomly dropped. Returns: Tensor: The tensor after DropBlock layer. )r$rr r r r rErr8r8r9forwardzDropBlock2d.forwardcCs*|jjd|jd|jd|jd}|S)Nz(p=z , block_size=z , inplace=))rG__name__rr r )rEsr8r8r9__repr__s&zDropBlock2d.__repr__Fr) rN __module__ __qualname____doc__floatintboolrCrrKstrrP __classcell__r8r8rFr9r@rs $ r@c sJeZdZdZddededededd f fd d Zd edefd dZ Z S) DropBlock3dz# See :func:`drop_block3d`. Frrr r r r Ncst||||dSrA)rBrCrDrFr8r9rCszDropBlock3d.__init__rcCrHrI)r=rr r r r rJr8r8r9rKrLzDropBlock3d.forwardrQ) rNrRrSrTrUrVrWrCrrKrYr8r8rFr9rZs$rZ)FrT)r torch.fxtorch.nn.functionalr functionalr+rutilsrrUrVrWr$r=fxwrapModuler@rZr8r8r8r9sR  2 5