Created by: ManWingloeng

paddlepaddle是自动判断权重共享的吗？我需要像tensorflow那样传reuse参数吗？

    def discriminator(self, x, y_, scope='discriminator', is_training=True, reuse=False):
        with tf.variable_scope(scope, reuse=reuse) :
            x = dropout(x, rate=0.2, is_training=is_training)
            y = tf.reshape(y_, [-1, 1, 1, self.y_dim])
            x = conv_concat(x,y)

            x = lrelu(conv_layer(x, filter_size=32, kernel=[3,3], 
                          layer_name=scope+'_conv1'))
            x = conv_concat(x,y)
            x = lrelu(conv_layer(x, filter_size=32, kernel=[3,3], stride=2, 
                          layer_name=scope+'_conv2'))
            x = dropout(x, rate=0.2, is_training=is_training)
            x = conv_concat(x,y)

            x = lrelu(conv_layer(x, filter_size=64, kernel=[3,3], 
                          layer_name=scope+'_conv3'))
            x = conv_concat(x,y)
            x = lrelu(conv_layer(x, filter_size=64, kernel=[3,3], stride=2, 
                          layer_name=scope+'_conv4'))
            x = dropout(x, rate=0.2, is_training=is_training)
            x = conv_concat(x,y)

            x = lrelu(conv_layer(x, filter_size=128, kernel=[3,3], 
                          layer_name=scope+'_conv5'))
            x = conv_concat(x,y)
            x = lrelu(conv_layer(x, filter_size=128, kernel=[3,3],
                          layer_name=scope+'_conv6'))
            x = conv_concat(x,y)

            x = Global_Average_Pooling(x)
            x = flatten(x)
            x = concat([x,y_]) # mlp_concat

            x_logit = linear(x, unit=1, layer_name=scope+'_linear1')
            out = sigmoid(x_logit)


            return out, x_logit, x

类似GAN中我需要重复使用D模型，paddle是直接判断shape然后判断是否reuse吗？