o #jq@s`ddlZddlmZddlmZddlmZmZm Z gZ ddd Z Gd d d Z dddZ dS)N) functional)Linear)Conv1DTransposeConv2DTransposeConv3DTranspose?cCs"tj|||jd}t|||S)N)shape)paddlenormalr assign)xmeanZstdZ temp_valuerc/var/www/html/Deteccion_Ine/venv/lib/python3.10/site-packages/paddle/nn/utils/spectral_norm_hook.pynormal_s rc@s:eZdZdddZddZd d Zd d Zed dZdS) SpectralNormweightr-q=cCs2||_||_|dkrtd|||_||_dS)NrzGExpected n_power_iterations to be positive, but got n_power_iterations=)namedim ValueErrorn_power_iterationseps)selfrrrrrrr__init__s zSpectralNorm.__init__csN|}jdkr|jgfddt|D}|jd}||dgS)Nrcsg|] }|jkr|qSr)r).0drrr 0sz9SpectralNorm.reshape_weight_to_matrix..)rZ transposeranger Zreshape)rr weight_matheightrr rreshape_weight_to_matrix*s  z%SpectralNorm.reshape_weight_to_matrixc Cst||jd}t||jd}t||jd}||}|rmtBt|jD]'}ttj tj ||dddd|j d|ttj t ||d|j d|q)|jdkr^| }| }Wdn1shwYt |t||}||}|S) N_orig_u_vTF)Z transpose_xZ transpose_yrZaxisepsilon)getattrrr&r no_gradr#rr F normalizematmulrclonedotmv) rlayerdo_power_iterationruvr$_sigmarrrcompute_weight7sH     zSpectralNorm.compute_weightcCst||j|j||jddS)N)r5)setattrrr:Ztraining)rr4Zinputsrrr__call__^s zSpectralNorm.__call__cCs>|jD]\}}t|tr|j|krtd|qt||||}|j|}t<| |} | j \} } | | g} t | dd} | | g} t | dd} t j| d|jd} t j| d|jd} Wdn1sjwY|j|j=||jd|t||j|d||jd| ||jd| |||S) Nz>Cannot register two spectral_norm hooks on the same parameter rr rr*r'r(r))Z_forward_pre_hooksitems isinstancerr RuntimeError _parametersr r-r&r Zcreate_parameterrr.r/rZ add_parameterr;Zregister_bufferZregister_forward_pre_hook)r4rrrrkhookfnrr$hwr6r7rrrapplyes6          zSpectralNorm.applyN)rrrr) __name__ __module__ __qualname__rr&r:r< staticmethodrFrrrrrs   'rrrrcCs:|durt|ttttfrd}nd}t||||||S)a Applies spectral normalization to a parameter according to the following Calculation: Step 1: Generate vector U in shape of [H], and V in shape of [W]. While H is the :attr:`dim` th dimension of the input weights, and W is the product result of remaining dimensions. Step 2: :attr:`n_power_iterations` should be a positive integer, do following calculations with U and V for :attr:`power_iters` rounds. .. math:: \mathbf{v} := \frac{\mathbf{W}^{T} \mathbf{u}}{\|\mathbf{W}^{T} \mathbf{u}\|_2} \mathbf{u} := \frac{\mathbf{W} \mathbf{v}}{\|\mathbf{W} \mathbf{v}\|_2} Step 3: Calculate :math:`\sigma(\mathbf{W})` and normalize weight values. .. math:: \sigma(\mathbf{W}) = \mathbf{u}^{T} \mathbf{W} \mathbf{v} \mathbf{W} = \frac{\mathbf{W}}{\sigma(\mathbf{W})} Refer to `Spectral Normalization `_ . Parameters: layer(Layer): Layer of paddle, which has weight. name(str, optional): Name of the weight parameter. Default: 'weight'. n_power_iterations(int, optional): The number of power iterations to calculate spectral norm. Default: 1. eps(float, optional): The epsilon for numerical stability in calculating norms. Default: 1e-12. dim(int, optional): The index of dimension which should be permuted to the first before reshaping Input(Weight) to matrix, it should be set as 0 if Input(Weight) is the weight of fc layer, and should be set as 1 if Input(Weight) is the weight of conv layer. Default: None. Returns: Layer, the original layer with the spectral norm hook. Examples: .. code-block:: python >>> from paddle.nn import Conv2D >>> from paddle.nn.utils import spectral_norm >>> paddle.seed(2023) >>> conv = Conv2D(3, 1, 3) >>> sn_conv = spectral_norm(conv) >>> print(sn_conv) Conv2D(3, 1, kernel_size=[3, 3], data_format=NCHW) >>> # Conv2D(3, 1, kernel_size=[3, 3], data_format=NCHW) >>> print(sn_conv.weight) Tensor(shape=[1, 3, 3, 3], dtype=float32, place=Place(cpu), stop_gradient=False, [[[[ 0.01668976, 0.30305523, 0.11405435], [-0.06765547, -0.50396705, -0.40925547], [ 0.47344422, 0.03628403, 0.45277366]], [[-0.15177251, -0.16305730, -0.15723954], [-0.28081197, -0.09183260, -0.08081978], [-0.40895155, 0.18298769, -0.29325116]], [[ 0.21819633, -0.01822380, -0.50351536], [-0.06262003, 0.17713565, 0.20517939], [ 0.16659889, -0.14333329, 0.05228264]]]]) Nrr)r>rrrrrrF)r4rrrrrrr spectral_normsE rK)rr )rrrN)r rr.Z layer.commonrZ layer.convrrr__all__rrrKrrrrs   o