deep-learning - stacked rnn (num layers > 1) 是如何在pytorch上实现的？

Question

pytorch 中的 GRU 层接受一个名为 num_layers 的参数，您可以在其中堆叠 RNN。然而，尚不清楚后续 RNN 究竟如何使用前一层的输出。

根据文档:

Number of recurrent layers. E.g., setting num_layers=2 would meanstacking two GRUs together to form a stacked GRU, with the second GRUtaking in outputs of the first GRU and computing the final results.

这是否意味着 GRU 第一层的最终单元的输出作为输入馈送到下一层？或者这是否意味着每个单元格(在每个时间步长)的输出作为输入馈送到下一层的同一时间步长的单元格？

Answer 1

Does this mean that the output of the final cell of the first layer of the GRU is fed as input to the next layer? Or does it mean the outputs of each cell (at each timestep) is fed as an input to the cell at the same timestep of the next layer?

后者。第一层每个时间步的输出被用作第二层相同时间步的输入。

这张图来自a Keras tutorial显示多层 RNN 的结构: