Open akaniklaus opened 5 years ago
I never tested it for LSTMs. The library itself should work for LSTMs out of the box. However, I am not sure if there might be problems with inducing too much sparsity in the recurrent weight matrices.
I will look into this and see if they are straightforward to use. If they are, I might add some LSTM starter code to try it quickly.
Would this also work for LSTM? Do you have any benchmark of the method on that?