Will soon release CuDNNGRU and CuDNNLSTM layers for Keras. A lot less features than the GRU and LSTM layers, but ~6-7x faster on GPU.
Show this thread
-
The first version won't feature dropout at all
The built-in dropout isn't helpful, and implementing something equivalent to recurrent dropout will be left to a later project
-
-
I think adding Recurrent Highway Connection would be great too
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.