当试图使用tf.cond()仅构造DropOutRapper操作时出错,如果DropOutRate>0

DropoutWrapper 如果指定 dropout_rate l2_alpha 乘数值大于零。

以下是我为辍学而努力实现的目标:

layers = []
for layer_size in self.hidden_sizes:
    cell = self.cell_type(layer_size)
    cell = tf.cond(
        self.dropout_rate > 0.0,
        lambda: tensorflow.contrib.rnn.DropoutWrapper(cell, output_keep_prob=(1.0 - self.dropout_rate)),
        lambda: cell
    )
    layers.append(cell)
stacked_layers = tensorflow.contrib.rnn.MultiRNNCell(layers)

这将引发错误:

`TypeError: Failed to convert object of type <class 'tensorflow.python.ops.rnn_cell_impl.DropoutWrapper'> to Tensor. Contents: <tensorflow.python.ops.rnn_cell_impl.DropoutWrapper object at 0x7fa51b84bbe0>. Consider casting elements to a supported type.`

有什么方法可以用 tf.cond() 或者其他方式?