对模型设计的一个疑问

[Xelawk]

为什么是这样：

        reconstructed_image = torch.tanh(im)  # [-1, 1]
        if self.long_skip:
            reconstructed_image = (reconstructed_image + synthesized).clamp(0,1)

而不是这样：

        if self.long_skip:
            reconstructed_image = reconstructed_image + synthesized
            reconstructed_image = torch.sigmoid(reconstructed_image)