o #j@s\ddlZddlmZddlmZmZmZddlm Z m Z ddl m Z gZ Gdd d e ZdS) N)_C_ops)core framework unique_name)_current_expected_placein_dygraph_mode) Initializercs*eZdZdZfddZdddZZS)Bilinearae This initializer can be used in transposed convolution operator to act as upsampling. Users can upsample a feature map with shape of (B, C, H, W) by any integer factor. Returns: Bilinear initializer instance objects. Examples: .. code-block:: python >>> import math >>> import paddle >>> import paddle.nn as nn >>> from paddle.regularizer import L2Decay >>> factor = 2 >>> C = 2 >>> B = 8 >>> H = W = 32 >>> w_attr = paddle.ParamAttr(learning_rate=0., ... regularizer=L2Decay(0.), ... initializer=nn.initializer.Bilinear()) >>> data = paddle.rand([B, 3, H, W], dtype='float32') >>> conv_up = nn.Conv2DTranspose(3, ... out_channels=C, ... kernel_size=2 * factor - factor % 2, ... padding=int( ... math.ceil((factor - 1) / 2.)), ... stride=factor, ... weight_attr=w_attr, ... bias_attr=False) >>> x = conv_up(data) Where, `out_channels=C` and `groups=C` means this is channel-wise transposed convolution. The filter shape will be (C, 1, K, K) where K is `kernel_size`, This initializer will set a (K, K) interpolation kernel for every channel of the filter identically. The resulting shape of the output feature map will be (B, C, factor * H, factor * W). Note that the learning rate and the weight decay are set to 0 in order to keep coefficient values of bilinear interpolation unchanged during training. cstdS)z$Constructor for BilinearInitializer.N)super__init__)self __class___/var/www/html/Deteccion_Ine/venv/lib/python3.10/site-packages/paddle/nn/initializer/Bilinear.pyr IszBilinear.__init__Nc Cs||}t|tjstdt|tjstd|j}t|dkr&td|d|dkr2tdtj t |jdd }|d}t |d }d|d |dd |}t t |D]"}||} |||} d t | ||d t | ||||<q]t||}|jtjjjtjjjtjjjfvrtjjj} |jtd d |jdg|j| tjjjdd} n|j} |} | tjjjkrd} dd|jD}ntd|jt |dkrtdtrt !| t"|| |t#|jtjjjtjjjtjjjfvr t $| |j}|%|dS| %|dS|j&dd| gid| dt"|| |id}|jtjjjtjjjtjjjfvrK|j&dd| id|i| j|jdd ||_'|S)!aZInitialize the input tensor with Bilinear initialization. Args: var(Tensor): Tensor that needs to be initialized. block(Block, optional): The block in which initialization ops should be added. Used in static graph only, default None. Returns: The initialization op zvar must be framework.Variable.zblock must be framework.Block.zthe length of shape must be 4.rz#shape[2] must be equal to shape[3].Zfloat32)dtypeg@r .Z bilinear_inittmpF)nameshapertypeZ persistableZ fp32_valuescSsg|]}t|qSr)float).0vrrr sz$Bilinear.forward..zUnsupported dtype %sizThe size of input is too big. NZ assign_valueZOutrr)routputsattrscastX)Zin_dtype out_dtype)rZinputsrr )(Z _check_block isinstancerVariable ValueErrorZBlockrlennpZzerosprodceilrangeabsZreshaperrZVarDescZVarTypeZFP16ZBF16ZFP64ZFP32Z create_varrgeneratejoinrZ LOD_TENSORZflat TypeErrorrrZ assign_value_listrr!Z_share_underline_tensor_toZ append_opop)rvarblockrweightsizefcixyr#Zout_varZ value_namevaluesZvar_tmpr1rrrforwardMs    .        zBilinear.forward)N)__name__ __module__ __qualname____doc__r r< __classcell__rrrrr s .r )numpyr(ZpaddlerbaserrrZbase.frameworkrrZ initializerr __all__r rrrrs