====== LSTM ======

  * https://pytorch.org/blog/optimizing-cuda-rnn-with-torchscript/

{{tag>RNN LSTM pytorch "temporal abstraction"}}