cookcodes
Results
2
issues of
cookcodes
In paper: "To achieve 40 ms look-ahead including the 10-ms overlap, we use M = 3" 40 ms look ahead, it means have only 30ms data to shift for comb...
Keras now use double bias, so compute_gru need to be adapted to it. see https://kaixih.github.io/keras-cudnn-rnn/ for more information